Bacterial strains used in this study are listed in Supplementary Table S1. Escherichia coli was grown in LB broth (LPS solution, Korea, Cat. LB-05) at 30°C or 37°C, depending on the replication origin (ori) sequence of the plasmids. E. coli DH5α was used as a cloning host for the construction of dual sgRNA plasmids. To introduce the D10A mutation into the cas9-NG gene, the E. coli HK1159 strain carrying the cas9-NG gene located downstream of the L-arabinose-inducible P _BAD promoter in the chromosome of E. coli MG1655 was used. pHK463 (Kim et al., 2020), which expresses λ Bet protein, was used for the introduction of the D10A mutation by a single-strand oligonucleotide. E. coli HK1159 carrying pHK463 plasmid was cultured in LB broth containing ampicillin at 30°C. When the OD_600nm reached 0.4, L-arabinose (TCI, Japan, Cat. A0515) was added at a final concentration of 1 mM to overexpress Cas9-NG nuclease and λ red Bet protein. After 3 h, the cultured cells were harvested, washed twice with 10% glycerol, resuspended, and stored at −80°C. pHL143 sgRNA plasmid (200 ng) and D10A mutagenic oligonucleotide (100 pmol) were electroporated into the pHK463/HK1159 cells under conditions of 25 μF, 200 Ω, and 1.8 kV using a 0.1 cm electroporation cuvette (Bio-Rad, United States, Cat. 1652089). Afterward, the cells were transferred to 950 μL of SOC medium immediately and recovered for 1 h at 30°C and 180 rpm. Afterward, the cells were spread on LB agar containing spectinomycin and incubated for 16 h at 37°C. The successful incorporation of the D10A mutation into the cas9-NG gene was confirmed via Sanger sequencing. After the pHK463 plasmid was cured at 42°C, the strain was designated as E. coli SH169, in which the cas9-NG(D10A) gene was located downstream of the L-arabinose-inducible P _BAD promoter in the chromosome. If needed, kanamycin (25 μg/mL), ampicillin (50 μg/mL), and spectinomycin (75 μg/mL) were added to the medium.

The sgRNA plasmids used in this study are listed in Supplementary Table S1. Primers used for plasmid construction are listed in Supplementary Table S2. To generate nicks on different strands of the double-stranded target DNA, a plasmid was constructed that simultaneously expresses two sgRNAs targeting different strands within the galK gene (Supplementary Figure S1). We designed the plasmid to enable stable replication by positioning the spectinomycin resistance gene and the replication origin of the plasmid between the two sgRNAs. This arrangement prevents recombination between identical nucleotide sequences in the sgRNA scaffolds. A DNA fragment containing the first P _j23119 -sgRNA gene from pTargetF plasmid (Addgene #62226) and the spectinomycin resistance gene and another DNA fragment containing the second P _j23119 -sgRNA gene and ori sequence of the plasmids were amplified by PCR. PCR was performed using KOD FX (ToYoBo, Japan, Cat. KFX-101). Two purified DNA fragments were ligated using Gibson Assembly Master Mix (NEB, United Kingdom, Cat. E2611) and transformed into DH5α competent cells. The sequences of the dual sgRNA plasmids were confirmed through Sanger sequencing.

Mutagenic oligonucleotides (Supplementary Table S3) were synthesized to introduce quadruple-bases substitution (⁴⁹⁴TTGT → GATC, ⁵⁰⁴TAAC → ATCA, or ⁵²⁸AATT → GTAG) or single-base substitution (C490T, T504A, or C523T), generating a premature stop codon in the target galK gene. Mutagenic oligonucleotide (500 pmol) and dual sgRNA plasmid (200 ng) were electroporated into the E. coli SH169 cells carrying the pHK463 plasmid, in which both Cas9-NG(D10A) nickase and the λ red Bet protein were overexpressed in response to the addition of L-arabinose. As a negative control, pSR017 plasmid, which contains dual sgRNAs targeting the galK (498–517) and xylB (638–657) genes, was used (Lim et al., 2023). Electroporation was performed as mentioned above. The recovered cells were spread on MacConkey agar (BD Difco, United States, Cat. 281810) containing D-galactose (0.5%) (Samchun Chemicals, Korea, Cat. G0476) and spectinomycin, or LB agar containing spectinomycin, and then incubated for 16 h at 37°C.

For single-nucleotide editing using PCR products as donor DNA, double-stranded donor DNAs were amplified using PCR from a strain in which the C490T, T504A, or C523T mutation was introduced into the galK gene. Mutagenic PCR products of 1 kb size (500 ng) and 5′-end 2-nucleotide (nt)-truncated dual sgRNA plasmid (200 ng) were electroporated simultaneously into the SH169 cells carrying the pKD46 plasmid (Datsenko and Wanner, 2000). Electroporated cells were spread on MacConkey agar containing D-galactose (0.5%) and spectinomycin.

Mutagenic oligonucleotide was synthesized to induce quadruple-bases substitution (⁴⁹⁴TTGT → GATC) in the galK gene, and the length of the oligonucleotides were 44mer, 67mer, 90mer, and 120mer. Mutagenic oligonucleotide (500 pmol) or pSH356 sgRNA plasmid (200 ng) was individually electroporated into L-arabinose-induced SH169 cells carrying the pHK463 plasmid and then incubated for 30 min at 37°C. Recovered cells were harvested, washed twice, and resuspended in a 10% glycerol solution. pSH356 sgRNA plasmid or mutagenic oligonucleotide was subsequently electroporated into each electrocompetent cell. After recovery for 30 min at 37°C, the cells were spread on MacConkey agar supplemented with D-galactose (0.5%) and spectinomycin.

Genome editing efficiencies for the galK target were calculated by counting the number of red and white colonies in MacConkey agar [white colonies/(white colonies + red colonies)]. To determine whether the target was cleaved by the CRISPR-Cas nickase system, the number of surviving colonies in MacConkey agar supplemented with D-galactose (0.5%) and spectinomycin or LB agar supplemented with spectinomycin were counted. Colonies were randomly selected, and the edited DNA targets were analyzed by restriction enzyme and Sanger sequencing to confirm the accuracy of the Cas9-NG(D10A) nickase-mediated genome editing.

For restriction enzyme digestion analysis, a 1 kb DNA fragment containing the target region was amplified from white colonies randomly selected from MacConkey agar supplemented with D-galactose using galK-F and galK-R primers. PCR products were then incubated with BclI restriction enzyme (Enzynomics, Korea, Cat. R048S) for 90 min at 37°C. The digestion products were confirmed through 2% agarose gel electrophoresis. Primers used for PCR amplification and Sanger sequencing are listed in Supplementary Table S2.

The sgRNA plasmids targeting the promoter of the xylA gene in the xylose operon were constructed. Various sgRNA plasmids with different lengths of target recognition sequence (TRS) were transformed into the SH169 cells, as mentioned above. Each of the single colonies in LB agar supplemented with spectinomycin was streaked on MacConkey agar supplemented with D-xylose (0.3%) and spectinomycin. In addition, the transformant cells were streaked on MacConkey agar supplemented with L-arabinose (0.3%) and incubated at 37°C for 12 h. If sgRNA/Cas9-NG(D10A) nickase complex can repress the xylA promoter, and cells are not able to metabolize D-xylose, the color of colonies remains white. Conversely, if cells can utilize D-xylose due to a failure to repress the xylA promoter, the color turns red.

The data collected from experiments were analyzed on Microsoft Excel 2016 (Microsoft Corporation, United States) using two-tailed unpaired t-test to evaluate significance.

We investigated the impact of the distance between nicks occurring around the editing site on the genome editing efficiency. A dual sgRNA plasmid was designed to have various distances between nicks (DBN), and a mutagenic oligonucleotide (120mer) was synthesized to introduce the four bases substitution (⁵⁰⁴TAAC → ATCA) in the galK gene. These constructs were electroporated into the E. coli SH169 cells overexpressing λ Bet protein and Cas9-NG nickase, a Cas9 nickase variant with expanded protospacer adjacent motif (PAM) (5′-NG) (Figure 1A; Supplementary Figure S2). When the target sequence in galK is properly altered by mutagenic oligonucleotides via the introduction of stop codons, the edited cells form white colonies in MacConkey agar containing D-galactose. Conversely, when the target was unedited, D-galactose can be normally metabolized, thereby forming red colonies (Supplementary Figure S3).

For the PAM-in design, in which the PAM sequence is located inside the double nicks, white colonies were not observed on MacConkey agar containing D-galactose, and the number of surviving colonies was high (≥10⁷ CFU/μg DNA). In contrast, for the PAM-out design, in which the PAM sequence is located outside the double nicks, white colonies were observed in the proportions of 9%, 11%, and 7% when the distances between the double nicks were 10, 20, and 26 base pairs (bp), respectively. When double nicks were formed 44 or 53 bp apart, the number of surviving colonies decreased significantly to 10⁵ CFU/μg DNA, and the proportion of white colonies was very high at 71% or 49%, respectively (Figures 1B, C). Quadruple-base substitutions in the galK target of randomly selected white colonies were confirmed through BclI restriction enzyme digestion and Sanger sequencing. As a result, for DBN 10, DBN 20, DBN 26, DBN 44, and DBN 53—where white colonies were observed—it was confirmed that the BclI restriction enzyme site had been inserted into the target sequence in four, four, four, two, and two out of four white colonies, respectively (Supplementary Figure S4).

Sanger sequencing revealed that all four white colonies were correctly edited in cases of overlapped targets (DBN 10, DBN 20, and DBN 26). However, in the case of DBN 44 and DBN 53, unwanted mutations were observed within the target sequence in all four colonies (Supplementary Figure S5). These outcomes suggest that while the probability of obtaining edited cells with high viability decreases when the bases to be edited are located on two overlapping target sequences (DBN 10, DBN 20, and DBN 26), the accuracy of the editing appears to be higher. Conversely, when the nucleotide sequence to be edited is positioned between targets, the target DNA is efficiently cleaved by the CRISPR-Cas9 nickase system, albeit with a decrease in editing accuracy. Additionally, these findings suggest that the outcomes of nickase-mediated genome editing cannot be conclusively verified using restriction enzymes. As a result, in subsequent experiments, the edited target sequences were analyzed through Sanger sequencing.

To enhance the genome editing accuracy under the PAM-out (DBN 44 bp) condition where target DNA is effectively cleaved, we examined how editing efficiency varies depending on the mutation locations. We designed mutations to be introduced on the target DNA hybridized with sgRNA or in between targets. Specifically, various lengths of mutagenic oligonucleotides (44mer, 67mer, 90mer, and 120mer) inducing four bases substitution at two mutagenesis sites (on Target1 and between Targets) were synthesized (Supplementary Figure S6). Each mutagenic oligonucleotide was electroporated along with a dual sgRNA plasmid with 44 bp DBN into cells overexpressing Cas9-NG(D10A) nickase and λ Bet protein.

In the case of On Target1, when L67-5′E, L90, and L120 were used, the percentage of white colonies was 96–97%, and almost all cells showed an edited phenotype. However, less than 5% of white colonies were obtained when L44 and L67-3′E were used (Figure 2A). In the case of Between Targets, when M67-5′E and M67-3′E were used, there was no significant difference in the proportion of white colonies, at 54% and 34%, respectively (Figure 2B). White colonies were randomly selected, and the nucleotide sequences of the edited galK targets were amplified and analyzed using Sanger sequencing (Supplementary Figure S7). When mutagenic oligos (L44, L67-5′E, L67-3′E, L90, and L120) that cause mutations in On Target1 were used, three, four, two, four and four out of four white colonies showed correct sequence editing, respectively. Conversely, when M67-3′E was used to induce mutations between targets, only one among four white colonies was correctly altered, while when M44, M67-5′E, M90, and M120 were used, all four white colonies showed unwanted mutations on the target DNA recognized by sgRNA. These results indicate that Cas9-NG nickase-mediated genome editing can be accurate when the nucleotide to be edited is located on the target sequence hybridized with sgRNA.

We investigated the genomic editing mechanism of the nickase system by employing a two-step electroporation method, which involves altering the order of electroporation for the dual sgRNA plasmid (pSH356) and the mutagenic oligonucleotides. As a result, when mutagenic oligos were electroporated first, followed by electroporation of pSH356 sgRNA plasmid, although the editing efficiency was lower than that achieved with simultaneous electroporation of both the oligo and sgRNA plasmid, the resultant patterns were similar (Figure 2A; Supplementary Figure S8A). However, when the sgRNA plasmid was electroporated first, white colonies were rarely obtained (Supplementary Figure S8B). These results indicate that oligonucleotides with homology arms are ineffective in facilitating the repair of cleaved chromosomes. When the target is cleaved, the majority of cells perish, making it impossible to obtain edited cells. Consequently, the results demonstrate that following mutagenesis, cells that have been edited can be successfully obtained via negative selection by employing the dual sgRNA/Cas9-NG nickase complex.

When PAM-in design sgRNA plasmid with a distance between double nicks at 48 bp apart was used, genome editing was not successful, and the number of survival colonies was increased, even if the DBN was sufficiently far (Figures 1B, C). To examine whether the PAM-in or PAM-out design, as well as the DBN, affects editing efficiency, we designed the PAM-out (46), which generates double nicks at a position similar to PAM-in (48) (Figure 3A; Supplementary Figure S9). In the case of PAM-in (48) dual sgRNA, when M120 oligo, which induces mutations (⁵⁰⁴TAAC → ATCA) between targets, was used, no white colony was formed, and the CFU was 10^6.9/μg DNA. When R120 oligo, which induces mutations (⁵²⁸AATT → GTAG) on the target, was used, the percentage of white colonies was low, at 3%, and the number of surviving colonies was similar to that of M120. In the case of PAM-out (46) dual sgRNA, when M120 oligo was used, the percentage of white colonies was 40%, and the number of surviving colonies was decreased to 10^4.6 CFU/μg DNA. When R120 oligo was used, the percentage of white colonies was remarkably increased to 94%, and the CFU was 10⁵/μg DNA (Figure 3B). Sanger sequencing was performed to confirm the editing accuracy (Figure 3C). When R120 oligo was used in PAM-in (48), three of the four white colonies showed correct sequence editing. In the case of PAM-out (46) using M120 oligo, only one among four white colonies was correctly edited. In contrast, with R120 oligo, only the desired mutation was introduced in all identified white colonies (Figure 3C). These results suggest that the PAM-out design effectively facilitates the formation of double nicks, and if base editing is designed on the target sequence hybridized by sgRNA, it is possible to achieve genome editing with significantly high efficiency.

We explored whether dual sgRNA with 20 nt-length sequences as TRS could be utilized for single-nucleotide editing by mutagenic oligonucleotide for each target at several mutagenesis sites. A mutagenic oligonucleotide that introduces a single-nucleotide substitution at three mutagenesis sites was electroporated with N₂₀ dual sgRNA (DBN 44 bp) into both Cas9-NG(D10A) nickase- and λ Bet-overexpressing cells. The proportions of white colonies obtained using mutagenic oligonucleotides that induce mutations in On Target1, Between Targets, and On Target2 were 71%, 72%, and 70%, respectively (Supplementary Figure S10A). Subsequently, the edited nucleotide sequences were verified using Sanger sequencing. However, unwanted mutations were observed in all identified white colonies (Supplementary Figure S10B). These data indicate that the accuracy of single-nucleotide editing was very low when dual sgRNA with a 20 nt target sequence was used.

To enhance the accuracy of single-nucleotide editing, we attempted to determine the length of the TRS in dual sgRNAs required to recognize a target and maintain the nickase activity. Based on previous studies (Lee et al., 2021; Lee et al., 2022), we expected that dual sgRNA with the minimum TRS length could exhibit mismatch intolerance between target DNA and sgRNA, leading to enhanced accuracy in single-nucleotide editing. After introducing dual sgRNAs, with truncations of 1–3 nt at the 5′-end, into Cas9-NG(D10A) nickase cells, we assessed the nickase activity in relation to the varied lengths of the 5′-end truncations in the sgRNAs. When dual sgRNAs with a 1 to 2 nt truncation at the 5′-end were transformed, the number of surviving colonies was approximately equal to the number of transformants containing untruncated dual sgRNA (10⁴ CFU/μg DNA). However, the CFU was increased to 10⁷/μg DNA when dual sgRNA with a 3 nt truncation was used (Figure 4A). These data indicate that the nickase activity of the sgRNA/Cas9-NG(D10A) nickase complex can be retained even when the 5′-end of the dual sgRNA is truncated by up to 2 nt. This implies that the truncation of 2 nt at the 5′-end of the dual sgRNA is necessary to enhance the accuracy of single-nucleotide editing (Figure 4B). Additionally, we determined the minimum number of nucleotides required for sgRNA binding to the target using the CRISPR interference system targeting the xylA promoter. Consistent with a previous study (Jeong et al., 2023a), we confirmed that gene expression can be repressed, even if the length of the TRS in sgRNA is shortened to 9 nt (Figures 5A–C). These results showed that in vivo, the 5′-end-truncated sgRNA/Cas9-NG(D10A) nickase complex can bind to a relatively short target sequence (∼9 bp), but the formation of nicks requires a minimum target length of 18 bp.

To confirm whether a single nucleotide can be efficiently edited using dual sgRNA with a truncation of 2 nt at the 5′-end, the dual truncated sgRNA plasmid was electroporated with mutagenic oligonucleotides introducing single-nucleotide substitutions at the three mutagenesis sites (Figure 6A). The percentages of single-nt-edited colonies obtained using oligonucleotides that induce mutations in On Target1 and On Target2 with the dual sgRNA truncated by 2 nt at the 5′-end were 13% and 29%, respectively. However, no single-nt-edited colonies were obtained in Between Targets (Figure 6B). The nucleotide sequences of the edited galK targets were analyzed using Sanger sequencing. In the case of On Target1 and On Target2, three and four of the four white colonies were correctly edited, respectively. In contrast, in the case of Between Targets, cells from four among four white colonies had unwanted mutations close to the targets (Figure 6C). These results indicated that nickase-mediated single-nucleotide editing could be performed using maximally 5′-end-truncated dual sgRNA and oligonucleotide introducing mutations on the target.

Additionally, genome editing was performed in cells overexpressing Cas9-NG nickase and λ Red using PCR products (∼1 kb) and 5′-end-truncated dual sgRNA plasmid (Supplementary Figure S11A). In the case of On Target1 and On Target2, the percentage of single-nt-edited colonies was 20% and 30%, respectively. In the case of Between Targets, no single-nt edited colonies were obtained on the MacConkey agar containing D-galactose (Supplementary Figure S11B). The accuracy of single-nucleotide editing was confirmed using Sanger sequencing (Supplementary Figure S11C). When the nucleotide to be edited is located on the target, four and four of four white colonies showed correct single-nucleotide substitutions (C490T and C523T), respectively. Conversely, when the nucleotide to be edited was Between Targets, the editing of a single nucleotide was not successful. These data show that Cas9-NG nickase-mediated single-nucleotide editing can be efficiently performed using dual sgRNA truncated by 2 nt at the 5′-end and donor DNA (oligonucleotides or PCR products) containing the nucleotide to be edited on the target sequence.

In addition, we calculated the proportion of single-nucleotide-edited colonies among 10 randomly selected colonies in LB agar. After electroporation of mutagenic oligonucleotide inducing a mutation in Target1 and 5′-truncated sgRNA into pHK463/SH169 cells, we randomly selected 10 candidate colonies that underwent single-nucleotide editing from LB agar containing spectinomycin (Supplementary Figure S12A). Sanger sequencing confirmed that three out of ten randomly selected colonies were accurately edited at a single-nucleotide level (Supplementary Figure S12B), consistent with the editing efficiency determined through phenotypic observation (Figure 6B; Supplementary Figure S11B).