Production of the en-AsCas12a (R1226A) variant and identification of the nicking property.
The CRISPR-Cas12a system consists of a single CRISPR RNA (crRNA) and Cas12a endonuclease that induces target DNA recognition15, 16. The Cas12a effector recognizes a T-rich protospacer adjacent motif (PAM) among target sequences (Fig. 1a, left inset), crRNA and target DNA hybridization induces an R-loop, and finally the active site of the RuvC domain (Fig. 1a, right inset) induces DNA cleavage17. The recently reported enhanced AsCas12a (en-AsCas12a) system shows that engineered amino acids near the PAM sequence (TTTN) dramatically enhances target DNA recognition, thereby increasing genome editing efficiency18. We constructed a nickase effector based on the en-AsCas12a (E174R, S542R, K548R) system, hereafter en-AsCas12a (R1226A), to effectively induce target-specific gene mutations with single- or dual mode of DNA targeting. To confirm the nicking property at the in vitro level, a cleavage assay was performed by targeting the plasmid containing the target (EMX1, CCR5) nucleotide sequence (Fig. 1b, c, Supplementary Fig. 1a, Supplementary Table 1, 2). As previously reported11, 12, the wild-type (WT) en-AsCas12a effector showed a minor nicking property while inducing DNA double strand break (Fig. 1c, Supplementary Fig. 1b). On the other hand, the en-AsCas12a (R1226A) form showed a typical activity as a DNA nickase by mainly inducing the open circular form of the plasmid (Fig. 1c). After confirming the nicking property of the en-Cas12a (R1226A) variant, we tried to determine whether en-Cas12a (R1226A) cleaves the target-strand or non-target strand on the target DNA. To this end, we co-treated the Cas12a (R1226A) variant with SpCas9 (D10A) or SpCas9 (H840A) nickases, whose cleavage points have already been identified, and the cleaved strand of target DNA by Cas12a (R1226A) variant was identified by the formation of double strand breaks (Fig. 1d). As a result, we found that when Cas12a (R1226A) variant, SpCas9 (D10A), and SpCas9 (H840A) nickases were simultaneously treated (Fig. 1e), double strand breaks were formed only when the Cas12a (R1226A) variant and SpCas9 (D10A) nickase were co-treated (Fig. 1f). Through these results, it can be seen that, like the wild-type Cas12a nickase, the en-Cas12a (R1226A) variant also mainly cleaves the non-target strand and induces nick formation on the target DNA (Supplementary Fig. 1c).
Optimization of the distance and direction between dual Cas12a (R1226A) modules to effectively induce mutations in target DNA
Based on these cleavage test results, target-specific mutations were induced on the target gene in a human-derived cell line (HeLa) using the dual-nickase-type en-AsCas12a (R1226A) or SpCas9 (D10A) effector (Fig. 2, Supplementary Fig. 2). Most of the induced mutation pattern shows large deletions, which is possibly generated by tandem nicking of sequential nickase binding (Supplementary Fig. 3, 6). To confirm the directionality issue of the CRISPR nickases, the mutation induction efficiency (%) was compared for the different combination according to the directionality of each nickase. For each combination of SpCas9 (D10A) and en-AsCas12a (R1226A) nickase, the indel frequency (%) in the target genes (EMX1, CCR5), which is induced in a PAM-out (Fig. 2a, c, e) and PAM-in (Fig. 2b, d, f) fashion, was analyzed by targeted amplicon sequencing (Supplementary Fig. 3, 6, Supplementary Table 3). Consequently, the indel frequency (%) induced in the direction of the PAM-out fashion (mean 12.4% for EMX1 and 5.8% for CCR5), (Fig. 2a, c, e) was higher than that in the PAM-in fashion (mean 0.9% for EMX1 and 1.0% for CCR5), (Fig. 2b, d, f) for all nickase combinations. Particularly, as the distance between the PAMs of nickase combinations increased, the overall indel frequency (%) decreased; a significant difference was observed depending on the nucleotide sequence in the target gene (EMX1, CCR5) and orientation between nickases (Fig. 2).
Comparison of the efficiency of mutation induction at target sites between dual- and single-Cas12a (R1226A) variants
The indel frequency (%) induced by the combination of dual nickases, each optimized with respect to PAM orientation, was compared with that induced by a single nickase or WT effector (Fig. 3). Most of the mutations formed by treatment with the single/dual nickase or WT effector were deletions (Supplementary Fig. 4, 6). The mutation frequency was highest for wt-enCas12a (mean 37.6% for EMX1 and 10.8% for CCR5), followed by dual-Cas12a (R1226A) (mean 6.6% for EMX1 and 3.5% for CCR5) and single-Cas12a (R1226A) (mean 2.1% for EMX1 and 0.6% for CCR5) (Fig. 3, Supplementary Fig. 3, 4, 6). As shown in Fig. 2, mutations were only effectively induced when binding in the PAM-out configuration was adopted using dual nickases (Fig. 3d–f, 3j, top). Notably, the targeted mutations formed by a single nickase were also observed for the dual nickase (Fig. 3, Supplementary Fig. 3, 4). The tendency of single nickase targeted mutation appears to be increased by the treatment of dual nickase rather than only single nickase treated (Fig. 3, Supplementary Fig. 3, 4, 6). Notably, when using the en-AsCas12a (R1226A) effector in a single nickase fashion, the editing efficiency (1.4–3.9%) was similar to that of the previously reported SpCas9 (D10A) nickase (0.4–2.2%), (Fig. 3d–f, j–l middle). Although the indel ratio (%) was lower than that for dual nickase, which is guided by two crRNAs, substantial indels (mean 2.1% for EMX1 and 0.6% for CCR5) were induced by intracellular delivery with the single nickase en-AsCas12a (R1226A) using one crRNA.
Target-specific genome editing with enhanced specificity using the dual-en-AsCas12a (R1226A) variant
To determine whether unintended mutations are induced (i.e., off-target effects) by the dual- and single-en-Cas12a (R1226A) variants, specificity was investigated for three target loci (DNMT1, AAVS1, and CCR5) (Fig. 4, Supplementary Fig. 5, 7). First, off-target candidates with up to 3 mismatched sequences compare to those of each target gene were selected based on an in silico analysis19, and targeted amplicons were prepared for each predicted off-target site (Fig. 4a, b, c, left). Then, the mutation frequency at each off-target site (%) was investigated by targeted amplicon sequencing (seeMethodssection, Supplementary Table 3). The target specificities (on-target editing (%)/off-target editing (%)) of the dual en-Cas12a (R1226A) variants were compared with that of the wild-type en-Cas12a effector, and histograms were obtained (Fig. 4a, b, c, right). In a comparison of three target loci, the wild-type en-Cas12a effector showed an editing efficiency of 73.4%, on average, compared with estimates of 31% for the dual en-Cas12a (R1226A) nickase. In comparison, for the selected off-target site, indel frequencies of 0.12% by dual en-Cas12a (R1226A) variants, compared to a frequency of 16% for the wild-type en-Cas12a effector (WT) (Fig. 4a, b, c, right). As a result, the target specificity was higher for the dual en-Cas12a (R1226A) variant (56.7-fold) than for the wild-type en-Cas12a effector at each genomic locus (Fig. 5).