Highly efficient and safe genome editing by CRISPR-Cas12a using CRISPR RNA with a ribosyl-2′-O-methylated uridinylate-rich 3′-overhang in mouse zygotes

The human DNMT1 gene was amplified through polymerase chain reaction (PCR) using Pfu polymerase (Biofact, PD302-50h). Primers were screened with Primer3web version 4.1.0 and synthesized by Bioneer Corporation (Korea). The amplicon and primer sequences for DNMT1 are noted in Fig.. The PCR products were purified with a HiGene Gel & PCR Purification system (Biofact, GP104-100), and cloned into an All in One PCR Cloning vector (Biofact, VT201-020). The cloned vector constructs were extracted using an NucleoBond Xtra Midi kit (Machery-Nagel, 740410.50). S1

Each gRNA was designed to target the DNMT1 or Trp 53 gene with a TTTV PAM (Fig. S1). Each gRNA was synthesized by Integrated DNA Technologies, Inc. For an in vitro cleavage assay, recombinant Acidaminococcus sp. Cpf1 (AsCpf1) was prepared. Codon-optimized Cpf1 CDS from Acidaminococcus sp. was ligated into a pET-28a(+) plasmid vector (Addgene, 69864-3), and the vector was cloned to transform BL21(DE3) E.coli cells (Thermo Fisher, EC0114). The transformed colony was cultured at 37 °C in LB broth in the presence of 50 µg/ml kanamycin following preculture for 4 h in the same medium. When the optical density reached ca. 0.6., isopropyl β-D-1-thiogalactopyranoside (IPTG) was added at 30 mM, and culture proceeded at 30 °C overnight. After induction, E.coli cells were harvested by centrifugation at 3000 rpm for 30 min. The pellets were suspended in phosphate-buffered saline (PBS) with an EDTA-free protease inhibitor cocktail and lysed in a sonicator for 30 min at 30% pulse power with 2-s intervals (Sonics Materials, VCX-500). The lysates were cleared by centrifugation at 13,000 rpm for 30 min. After being filtered through a 0.22 µm filter unit, the supernatant was loaded onto a Ni²⁺-affinity column using an FPLC Purification System (ÄKTA Purifier, GE Healthcare). The concentrations of the purified proteins were determined from Coomassie blue staining of the bands on SDS-PAGE gels using bovine serum albumin (BSA) as a calibrator.

Target-carrying plasmid vectors were digested in the presence of AsCpf1 protein (2.5–5 ng/µl) and 10 pM gRNA in 1X NEB 3.1 buffer. After digestion at 37 °C for 1 h, each sample was loaded onto a 1% agarose gel containing 5% ethidium bromide (EtBr). The band intensities were measured by illumination with ultraviolet (UV) light to calculate the levels of DNA cleavage.

C57BL/6J mice were cared for and used for preparation of fertilized eggs. The procedures for the use and care of mice were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC), KRIBB. All mice were bred in an isolated facility that was specific pathogen-free with a constant temperature of 24 °C, a humidity of 40%, and light cycles of 12 h. Forty-eight hours before human chronic gonadotropin (hCG) (Sigma-Aldrich, CG100) injection, 5 IU of pregnant mare serum gonadotropin (PMSG) (Daesung Microbiological Labs, A7101) was intraperitoneally injected into 5-week-old C57BL/6J female mice. Two days after mating, zygotes were collected from the ampullae of the oviducts, and the cumulus cells surrounding the zygotes were degraded by incubation in M2 medium (Sigma-Aldrich, M7167) containing 3 mg/ml hyaluronidase (Sigma-Aldrich, H3506) at room temperature for 5 min. After several washes in M2 and KSOM media (Merck, MR-121-D), the zygotes were cultured in KSOM medium for 5–6 h to select fertilized eggs. Fertilized eggs that formed polar bodies were used for RNP delivery.

Fertilized eggs were suspended in Opti-MEM (Gibco, 31985-070) in the presence of RNP and placed on microscope slides with platinum-plated electrodes (NEPA GENE, CUY505P5). RNP was delivered into zygotes using a NEPA 21 Super Electroporator (NEPA GENE) according to the recommended protocol. Four days after incubation of the eggs in KSOM medium at 37 °C, cells at the morula stage were harvested for preparation of genomic DNA.

Morula cells were suspended in PCR tubes with blastocyst lysis buffer (BLB) containing 60 µg/ml yeast tRNA (Invitrogen, AM7119) and 125 µg/ml proteinase K (Qiagen, 1114886). To prepare crude genomic DNA, the PCR tubes were placed in a thermal cycler, incubated at 95 °C for 10 min and 56 °C for 20 min and then stored at 4 °C. Using a REPLI-g Mini Kit (Qiagen, 150025), the crude genomic DNA was PCR-amplified using KAPA HiFi HotStart DNA Polymerase (Roche) according to the manufacturer’s instructions. The resulting PCR amplicons carrying Illumina TruSeq HT dual indexes were subjected to 150-bp paired-end sequencing using an Illumina iSeq 100. The indel frequencies were calculated by MAUND, which is available at https://github.com/ibs-cge/maund↗.

To compare the specificity of each gRNA, potential off-target sites associated with a Trp53 on-target sequence were screened with the Cas-OFFinder program (http://www.rgenome.net/cas-offinder↗) with criteria of fewer than two bulges and three mismatches.

Statistical tests of indel efficiency were performed in SigmaPlot software using a two-tailed Student’s t test. P-values < 0.05 were considered to indicate significance and are presented in the Legends section.

A variety of chemical modifications have been extensively explored for Cas9 and Cas12a in the directed repeat and spacer regions of the guide RNAs. However, no experimental efforts have been made to investigate the effects of chemical modifications in the U-rich 3′-overhang region in crRNA on the functionality of the Cas12a system. Based on previous reports that modifications in bases or ribose sugars of guide RNAs improve stability and editing efficiency^20–33, we focused on chemical modifications of bases and ribose sugars in the U-rich regions (Fig. 1a). The base modifications into nitrogen (N₅) were made at C₅ in the pyrimidine ring to render a pseudouridine base (ψ). The hydrogen atom attached to C₅ was also replaced with a methoxyl group, thereby producing 5′-methoxyuridine (moU). Alternatively, the hydroxyl group at C2 in ribose was modified with a fluorine or methyl group, hereafter designated F or OM, respectively (Fig. 1b).

We employed an in vitro digestion assay to assess guide RNA efficiency, in which plasmid DNA vectors harboring a 20-nt DNMT1 protospacer sequence were used as digestion substrates for the Cas12a-crRNA RNP complex (Fig. 2a). Under limited-digestion conditions, a subset of supercoiled plasmid vectors was transformed through an open circular state into linearized DNA. The ratio of linear to supercoiled forms and/or linear to open circular forms was regarded as a measure of crRNA efficiency. The incubation time and concentrations of RNP and substrate plasmid DNA were predetermined in this experiment. In line with the findings of the previous study, the U-rich crRNA significantly increased the DNA digestion activity in vitro (lane 4 vs lane 5). However, base modification of the crRNAs (ψ and moU) did not yield any increased cleavage activity (Fig. 2b). Rather, it compromised the modulatory effects of the U-rich 3′-overhangs. The F U-rich crRNA was also ineffective in modulating crRNA efficiency. In contrast, the OM U-rich crRNA and the unmodified U-rich crRNA showed comparable levels of digestion efficiency. Thus, we decided to test OM-crRNA under altered conditions. We increased the input of AsCas12a in the presence of equal amounts of crRNA and plasmid DNA but used a shortened incubation time (Fig. 2c). As the input of AsCas12a increased, a difference in cleavage activity became evident between unmodified and OM U-rich crRNA (p < 0.05). Notably, OM U-rich crRNA showed markedly higher efficiency than canonical crRNA (Fig. 2d). Taken together, these findings indicate that ribosyl-2′-O-methylation in the 3′-overhang of crRNA further potentiated the efficiency of Cas12a in vitro.

In addition to efficiency, toxicity is a critical issue in genome editing in zygotes because the number of zygotes available is limited. Although modification of the 5′-terminal triphosphates of guide RNAs into hydroxyl groups has been shown to improve cellular viability²⁶, less is known about the effects of OM modifications, particularly in the U-rich 3′-overhang, on the viability of zygotes. To investigate the cellular toxicity of OM U-rich crRNA, we injected AsCas12a with either canonical or modified crRNAs into zygotes and monitored viability at the morula stage 4 days after injection (Fig. 3a). For this, we tested the toxicity of RNP preparations in zygotes in five independent experiments. The first three experiments (experiment 1 to experiment 3) yielded, overall, low survival rates for all the types of crRNAs. The overall low survival rates likely stemmed from the long durations of zygote manipulation due to the large number of zygotes tested (n = 50, 66, 62 for each type crRNA). Nonetheless, OM crRNA enabled high-viability genome editing in zygotes (Fig. 3b). To shorten the duration of zygote manipulation, we decreased the number of zygotes to 22 and 24 for experiments 4 and 5, respectively. As a result, the overall survival rate increased significantly to more than 60%. In these two experimental batches, OM U-rich crRNA was responsible for the improved viability of zygotes (Fig. 3b, Table S1). To investigate whether OM U-rich crRNA consistently shows reduced toxicity for multiple targets, we added two additional murine genes, Wnk1 and Ahrr, and investigated the cellular toxicity in murine zygotes. The results indicated that OM U-rich crRNA versus canonical crRNA was responsible for the higher viability of murine zygote cells (Fig. 3c, Table S1, S2, and S3). Thus, we concluded that U-rich crRNA with ribosyl-2′-O-methylation lowered cellular toxicity when injected into zygotes as an RNP complex. Currently, it is unclear why ribosyl modifications lower cellular toxicity, but it is tempting to speculate that the 2′-hydroxyl group in RNA is chemically reactive and that methoxyl modification might prevent unwanted intracellular molecular reactions that may lead to cellular toxicity.

Then, we investigated whether OM U-rich crRNA shows improved genome editing efficiency in zygotes as measured by a DNA cleavage assay in vitro. Because experiments 4 and 5 showed convincingly high cell viability, we investigated indel efficiency in morula cells of the two experiments. The viable morula-stage cells were pooled, and genomic DNA was prepared. Deep sequencing analysis was performed with PCR-amplified DNA products, with Trp53 as the target. As compiled in Table 1, the U-rich crRNA significantly improved indel efficiency (0.296 to 0.783 and 0.173 to 0.979 for experiments 4 and 5, respectively). OM crRNA either retained or further improved the indel efficiency of the U-rich crRNA-AsCas12a RNP complex. Notably, in experiment 4, all 18 surviving morula cells showed biallelic indel mutations (Table 2, Fig. S2). In experiment 5, OM U-rich crRNA also showed a better indel efficiency than a canonical crRNA. Although the OM crRNA was slightly less efficient than an unmodified U-rich crRNA, the overall yield of mutations was higher than that of the unmodified U-rich crRNA considering the higher viability. Because the increased indel efficiency of OM U-rich crRNA was similarly observed for the Wnk1 and Ahrr genes, we conclude that OM U-rich crRNA facilitates the generation of genome-engineered mouse models with increased safety and efficiency. Further study is needed to determine whether the improved functionality of the engineered crRNA is applicable to zygotes of animals other than mice.

Despite the improved indel efficiency and mitigated cellular toxicity of OM U-rich crRNA, there remained a concern that the engineered guide RNA might impair the innately high specificity of Cas12a, thereby compromising the utility of the engineered CRISPR–Cas12a system. To investigate this possibility, we measured off-target effects in a targeted manner. We searched for potential off-target sites for Trp53 using the Cas-OFFinder program with criteria of fewer than three mismatches and two bulges. We screened five potential sites (Table) and performed deep sequencing analysis using genomic DNA prepared from the morula-stage cells of the previous experiments 4 and 5. S4

As shown in Table 2, OM U-rich crRNA did not significantly affect the off-target activity for four (OF1, OF2, OF3, and OF5) of the five potential off-target sites (Table 2). It appeared that OF1 showed a high indel frequency independent of the type of crRNA. However, we found that it originated from sequencing errors because OF1 has a polyguanylated sequence immediately downstream of the protospacer sequence. We manually investigated the sequencing data and found that, apart from the poly-G motif, few indel mutations were found for the three crRNAs tested. One intriguing result was observed for OF4, which carried a sequence almost identical to that of the target site with a single nucleotide mismatch. Moreover, the mismatch resided in an area with a high mismatch tolerance in the CRISPR-Cas12a system^34,35. For OF4, compared to canonical and unmodified U-rich crRNA, OM crRNA increased the level of indel mutations in one experimental batch (experiment 5). We hypothesize that the increased indel frequency may have originated from the improved efficiency of the modified crRNA. It is important to mention that a more careful identification of potential off-target sites is necessary, and accordingly, targets with off-target sites showing high sequence similarities should be avoided through in silico analysis. Then, highly specific and efficient genome engineering can be achieved via 3′-U-rich crRNA with ribosyl-2′-O-methylation in the CRISPR-Cas12a system.