The B. subtilis strains and plasmids used in this study are listed in Table 1. E. coli MC1061 was used to construct recombinant plasmids. B. subtilis cells were cultured in Luria–Bertani (LB) broth or LB agar (Difco, Detroit, MI, United States) media at 37°C for screening and flow cytometry measurement. When required, the antibiotics were supplemented with following conditions for selection purposes: ampicillin (100 μg/ml), chloramphenicol (5 μg/ml), or neomycin (10 μg/ml). Transformation of B. subtilis cells was carried out by the two-step transformation procedure (Cutting and Harwood, 1990).

The competent cell of B. subtilis 168 prepared by the previous method (Kim et al., 2015) was transformed with SpCas9-carrying plasmid pHCas9 to construct BS-C100. Next, BS-C100 was transformed with the pAD123 derivatives containing donor DNA template and sgRNA transcription module. Half of the transformation mixture was spread on LB agar plate containing chloramphenicol (5 μg/ml) and neomycin (10 μg/ml). The other half was inoculated into same conditioned broth media with final 1% concentration for prolonged incubation. After culturing for different time periods, the cultured cells were serially diluted and spread on LB agar plate containing chloramphenicol (5 μg/ml) and neomycin (10 μg/ml). The mutations were confirmed by PCR, phenotypic screening, flow cytometry, and DNA sequencing.

Flow cytometry and data analysis were conducted using a Guava^® easyCyte cytometer (Millipore Corporation, Billerica, MA, United States). A Guava InCyte assay demonstrated the GFP activity of B. subtilis 168 and BS-C111 cells. Bacillus strains were pre-cultured overnight and main-cultured the next day in LB medium. All the samples were washed with phosphate-buffered saline (PBS) three times and resuspended in 1 ml of PBS for further flow cytometry analysis. Each sample went through 15000 cell counts for 600 s, and was mixed for 3 s at a medium rate.

It has been established that 70 bp were required for RecE-dependent homologous recombination in Bacillus (Khasanov et al., 1992). In genome engineering using mazF counter-selectable marker, 125 bp-directed repeats for in vivo recombination were used to generate bpr deletion mutant (Zhang et al., 2006). To determine the size of the donor DNA template for HDR-mediated CRISPR-Cas9 in B. subtilis, we targeted the spo0A gene. Spo0A is a transcription regulator for initiation of sporulation in B. subtilis. Inactivation of spo0A prevents endospore formation and alters Bacillus cell morphology leading to cell lysis (Hamon and Lazazzera, 2001). Using two plasmids: pHCas9 and pB0A, we determined the appropriate size of the donor DNA for CRISPR-Cas9-mediated genome engineering. SpCas9-expressing plasmid pHCas9 (Figure 1A) was transferred into wild type B. subtilis 168 to give strain BS-C100. Prior to introducing the donor DNA template plasmid pB0A (Figure 1B), we transfer p0AgR, which contains only a spo0A-targeting sgRNA transcribing module, into BS-C100. Introduction of p0AgR produced no colonies on LB chloramphenicol (5 μg/ml) and neomycin (10 μg/ml) agar plate (data not shown). This result indicates that p0AgR was unable to generate the spo0A deletion mutant without a donor DNA template for homologous recombination.

To determine the minimum size of donor DNA template that retains high recombination efficiency we truncated the donor DNA template from 1 kb to 50 bp for each upstream and downstream (Figure 2A). The truncated donor DNA templates were cloned into spo0A-targeting sgRNA producing plasmids and produced pB0A-2, pB0A-1, pB0A-800, pB0A-400, pB0A-200, and pB0A-100. Introduction of each plasmids into BS-C100 resulted in strains containing a 32 bp deletion mutation in spo0A: BS-C101, BS-C102, BS-C103, BS-C104 BS-C105, and BS-C106. With a 2 kb donor DNA template (1 kb each from the N- and C-termini), mutation efficiency was 100%. However, with truncated donor DNA templates, from 1 kb to 400 bp, the mutation efficiencies were slightly lower as 91% on average (Figure 2B). With a 200 bp donor DNA template, there was a substantial decrease in mutation efficiency, which indicates that the donor DNA template should be longer than 400 bp for maintaining the high mutation efficiency using HDR-mediated CRISPR-Cas9 in B. subtilis. We also observed that mutation efficiency was very low without the pHCas9, as 8.7% even with a 2 kb donor DNA template. This suggests that the mutation generation strongly depends on SpCas9 (Figure 2B). The spo0A mutants were distinguished by morphological change (Figure 2A) and confirmed with DNA sequencing.

To determine whether the size of the deletion region affects mutation efficiency in B. subtilis, we constructed 500 bp and 1 kb deletions in the spo0A. It has already been demonstrated that the oligonucleotide-mediated HDR had 30 and 6.5% mutation efficiency in Lactobacillus with, respectively, 501 and 1,002 bp deletions on the lacL gene (Oh and van Pijkeren, 2014). We obtained strains BS-C107 and BS-C108 by introducing pB0A-d500 and pB0A-d1000 into BS-C100. The 500 bp deletion mutant strain BS-C107 showed 97% mutation efficiency and the 1 kb deletion mutant strain BS-C108 showed 88% mutation efficiency (Figure 2C). Unlike in Lactobacillus, there was no significant decrease in mutation efficiency among three different spo0A mutants, BS-C102, BS-C107, and BS-C108. This result indicates that plasmid delivered SpCas9/sgRNA, along with 500 bp donor DNA template maintains high mutation efficiency for the deletion up to 1 kb in B. subtilis.

Previous large genomic deletion studies have reported low mutation efficiency (Jeong et al., 2015) or used prerequisite strains containing antibiotic resistance markers (Morimoto et al., 2009). To determine whether the CRISPR-Cas9 system can efficiently delete an operon-like large genomic region in the chromosome and leave it completely free of foreign DNA, we targeted the pps operon. Research using a minimal B. subtilis genome factory confirmed that the plipastatin-synthesizing 38 kb operon (ppsABCDE) is not essential for cell growth (Ara et al., 2007). First, we constructed a ppsC targeting sgRNA, which generates DSB on the middle of 38 kb pps operon. A donor DNA template, which contains the N- and C-termini of the pps operon was PCR amplified and cloned into the ppsC targeting sgRNA-generating plasmid. Then we introduced the resulting plasmid into pHCas9-containing BS-C100 and cultured on LB chloramphenicol (5 μg/ml) and neomycin (10 μg/ml) agar. However, the transformants failed to generate the entire operon deletion (data not shown). We suspected that the DSB site was generated too far from the each N-and C-termini recombination region to initiate an efficient double cross-over. Therefore, we designed two sgRNAs, which can produce DSBs at each end of pps operon near the HDR sites, to yield a precise large genomic deletion: one targeting the N-terminus of the ppsA and the other targeting C-terminus of the ppsE (Figures 3A,B). After complexing with SpCa9, ppsA- and ppsE-targeting sgRNA will generate DSBs on the N-and C-termini of the pps operon respectively.

We introduced pBP-dpps into BS-C100 and demonstrated whether the two DSBs, which completely eliminates the 38 kb in the B. subtilis chromosome, can yield an efficient mutagenic HDR (Figure 3C). The pBP-dpps transferred BS-C100 produced a strain BS-C109, and the newly obtained BS-C109 transformation reaction mixture was spread onto agar containing LB chloramphenicol (5 μg/ml) and neomycin (10 μg/ml). Transformation efficiency was 3.13 × 10² transformants/μg DNA. To confirm deletion of the pps operon we used two primers, pps-F1 and pps-R2, and amplified 3′-dacC through 5′-yngK. In the wild type strain, this region is over 40 kb, which exceeds the maximum amplification size under our PCR condition. However, 38 kb of this 40 kb region is absent from the ppsABCDE deletion mutant and the remaining 2 kb can be amplified for the confirmation of the deletion mutation (Figure 3C). We obtained 217 transformants and 12 single colonies were randomly chosen for PCR amplification. Only 2 out of the 12 colonies (16.7%) were identified as pps operon deletion mutants (Figure 3D). The CRISPR-Cas9-mediated large genomic deletion achieved higher mutation efficiency than the synthetic gene circuit method (6.4%), but still had low mutation efficiency compared with the other CRISPR-Cas9-based mutations in B. subtilis.

To increase the operon-like large genomic deletion efficiency, we developed an iterative system based on the plasmids used in this study. Two plasmids, which produce SpCas9, spo0A-targeting sgRNA, and donor DNA template, were stably maintained under selective pressure. For this reason, when the target chromosomal region escapes sgRNA/SpCas9 endonuclease activity, it will “iteratively” attack the target region to form DSBs until mutagenic repair occurs (Figure 4A).

To account for the iterative genome engineering system using CRISPR-Cas9, we spread half (100 μl) of the newly obtained BS-C109 transformation reaction mixture onto the agar containing LB chloramphenicol (5 μg/ml) and neomycin (10 μg/ml). The other half (100 μl) of the BS-C109 transformation mixture was cultured for 16 h on LB chloramphenicol (5 μg/ml) and neomycin (10 μg/ml) broth media. The mixture was then spread in the same agar medium. Without the prolonged incubation, only 16% of the transformants were confirmed as pps operon deletion mutants. Surprisingly, after prolonged incubation, 16 out of the 20 colonies (80%) were identified as pps operon deletion mutants (Figure 3E). This result provided an evidence that prolonged incubation increased the mutation efficiency.

To confirm the effect of prolonged incubation on increasing the mutation efficiency, we introduced pB0A-1 and pB0A-d500 into BS-C100 respectively, and obtained strains BS-C102 (32 bp spo0A deletion) and BS-C107 (500 bp spo0A deletion). Each transformation reaction mixture was time-series incubated under antibiotic pressure. Without prolonged incubation (0 h), BS-C107 and BS-C102 transformants showed average mutation efficiencies of 83 and 85% respectively. Incubation for longer periods resulted increase in mutation efficiencies up to 100% in time dependent manner on both strains (Figure 4B). These results demonstrated that the mutation efficiency can be maximized by maintaining SpCas9, sgRNA transcription module and donor DNA template containing plasmids under selective pressure.

We targeted sigE to determine whether our system can produce genomic point mutation efficiently. Mutation in sigE gene is known to inactivate the sporulation cascade causing morphological change in B. subtilis (Stragier et al., 1988). By changing thymine (T) into adenine (A), we created a stop codon (TAA) at the N-terminus of sigE to prevent its translation, which causing a significant morphological change in the cell (Figure 5A). Transfer of sigE point mutation-generating plasmid pBE-pm into BS-C100 resulted a strain BS-C110, and half of the fresh transformation reaction mixture (100 μl) was spread onto agar containing LB chloramphenicol (5 μg/ml) and neomycin (10 μg/ml). The initial transformation efficiency was 2.84 × 10² transformants/μg DNA. On average, 33% of transformants (56 out of 169) showed a sigE mutant phenotype in this medium. Ten transparent colonies were randomly selected and sequence analysis confirmed that all 10 colonies contained sigE point mutation (T → A). We then cultured the other half of the transformation reaction mixture (100 μl) overnight in the same medium, under the same conditions. Spreading 100 μl of 10^-5-diluted BS-C110 overnight culture produced an average of 805 colonies and about 68.5% of mutants were shown to be sigE mutants (Figure 5B). Among the transparent colonies (Figure 5C), 10 colonies were randomly selected for DNA sequencing. Sequence analysis confirmed that all 10 colonies contained sigE point mutation (T → A). This result indicates that given prolonged incubation helped to increase genomic point mutation efficiency.

As well as generating knock-out mutants, we determined the effect of our iterative genome engineering method on generating knock-in mutants in B. subtilis. We designed pBE-gfp, which generates a GFP insertion mutant on the sigE gene. Introducing pBE-gfp into pHCas9-containing BS-C100 resulted in BS-C111 strain (Figure 6A). We spread half (100 μl) of newly obtained BS-C111 transformation reaction mixture onto LB agar containing 1% xylose in addition to chloramphenicol (5 μg/ml) and neomycin (10 μg/ml), and the transformation efficiency was 2.04 × 10² transformants/μg DNA. Among 202 single colonies, 135 colonies (66.8%) had a sigE mutant phenotypes with GFP expression. The other half (100 μl) of transformation reaction mixture was cultured overnight in the same medium at 37°C 200 rpm. As a control, B. subtilis 168 was cultured in LB without any antibiotics. Spreading 100 μl of 10^-4-diluted BS-C111 overnight culture yielded 442 distinct colonies of which 428 sigE mutants showed GFP expression (96.8%). Six of the 428 colonies were randomly selected for PCR amplification with primers sigE-NF1 and sigE-CR1. We verified the insertion of a P_xyl-GFP cassette using PCR and the sigE mutant-specific morphological change (Figure 6B) as well as DNA sequencing. Guava InCyte assay confirmed the GFP expression of BS-C111, while no GFP expression observed with the wild type strain B. subtilis 168 (Figure 6C). The iterative genome engineering method with prolonged incubation under selective pressure had also confirmed its ability to increase the insertion mutation efficiency.

For industrial and food-grade applications, traces of foreign DNA in the cells must be removed by plasmid curing. Previous research had demonstrated that the B. subtilis integration vector pHT01 and expression vector pAD123 have 50 and 43% stability under overnight antibiotic-free conditions, respectively (Yamashiro et al., 2011). To confirm the curing efficiency of plasmid pHT01 derivative pHCas9 containing a neomycin-resistance gene, BS-C101 was cultured overnight in 1 ml LB chloramphenicol (5 μg/ml) broth medium without neomycin. Single colonies were isolated by serial dilution and 70 colonies were randomly duplicated on agar containing LB chloramphenicol and LB neomycin. Sixteen out of 70 colonies did not grow on LB neomycin-containing plates, indicating that 22.8% of the Cas9-containing pHT01 derivative pHCas9 had been removed. One of the 16 colonies from the LB chloramphenicol-containing replica plate was cultured in 1 ml of LB medium to determine the curing efficiency of the sgRNA and donor DNA template-containing pAD123 derivative, pB0A. Duplicate samples of 70 randomly selected colonies were transferred onto agar plates with or without LB chloramphenicol (5 μg/ml). Thirty-five out of 70 colonies were not able to grow on agar containing LB chloramphenicol (5 μg/ml), indicating that both plasmids were easily cured by serial culture in antibiotic-free conditions (data not shown).