Previously, we isolated a B. pumilus wild-type strain from soil, and the genotype was the same as that of B. pumilus SH-B9 (NCBI Accession no: NZ_CP011007.1) according to whole-genome sequencing. The other strains used in this study are listed in Supplementary Table S3. Escherichia coli DH5α (the host for pCas9) and Escherichia coli Top10, both purchased from Thermo Fisher Scientific (Waltham, MA, United States), were used for plasmid construction and amplification, and cultured in Luria-Bertani medium containing 25 μg/ml chloromycetin. All chemicals, a DNA gel purification kit, and a plasmid extraction kit were purchased from TIANGEN Biotech (Beijing, China). Restriction enzymes and DNA ligase were purchased from New England Biolabs (Ipswich, MA, United States). Taq DNA polymerase and oligonucleotides used for PCR experiments were purchased from Ruibio Biotech (Beijing, China).

The pCas9 plasmid and sequences used in this study (Supplementary Table S4) are available from the Addgene nonprofit plasmid repository (Cat: 42,876). The pCas9 plasmid is 9,326 bp based on the low-copy backbone of pACYC184 (ATCC 37033), which endows resistance to chloramphenicol (25 μg/ml). The other elements of this plasmid include tracrRNA, Cas9, and an array of crRNAs with two BsaI sites for spacers (Figure 1). The plasmids pCas9-Cax, pCas9-Cac and pCas9-Cpt were constructed for cre-box editing of seven target genes. The spacer oligonucleotides for cloning into pCas9 and the primers used for PCR are listed in Supplementary Table S5.

For editing multiple cre-box sites at the seven operons and simultaneously controlling the range of gRNA:Cas9 complex off-targets, oligonucleotides of gRNAs must direct the Cas9 protein to recognize multiple target sites. Because the cre-boxes of B. pumilus have the consensus sequence WWGNAANCGNWNNCW (Supplementary Table S1), we chose NNG for the PAM and cre proximal sequences for target sequences designed using the Zhang Lab website (http://crispor.tefor.net/), and the results (Supplementary Table S2) include partial or complete cre sequences (Supplementary Figure S2). In order to better describe the seven target operons, we continued using the same gene symbols of the type strain B. subtilis168 (NCBI Accession no: NC_000964.3) where annotation was consistent. The seven target sequences were divided into three groups with the highest (70%), moderate (70–50%), and lowest (50–30%) identity with the three unspecific gRNAs (Cac, Cpt, and Cax).

Among them, ackA and ntdA share the highest identity (70%) with the unique gRNA Cac and the common gRNA Cpt , both of which share >50% identity with ptsH and budA. The ackA encoding acetate kinase that converts pyruvate into acetate and is positively regulated by CcpA protein, ntdA encoding an aminotransferase responsible for converting dehydroglucose-6-phosphate to kanosamine-6-phosphate, ptsH encoding phosphocarrier protein Hpr that is also repressed by CcpA, and budA encoding acetolactate decarboxylase that converts pyruvate to acetoin and is also activated by CcpA. The target site sharing only 30% identity is the sucC operon encoding succinyl-CoA synthetase, which converts succinyl-CoA to succinate in the tricarboxylic acid (TCA) cycle and is negatively regulated by CcpA, and shares the gRNA Cax with two other operons (acsA and xylA) but they share more than 50% identity. Acetyl-CoA synthase, encoded by acsA, converts pyruvate into propanoate, which can be activated by CcpA and xylA encoding a xylose isomerase responsible for converting D-glucose into D-fructose. The three unspecific gRNAs (Cac, Cpt, and Cax) were inserted into the BsaI sites of pCas9 as the CRISPR spacers after annealing at 95°C for 5 min and 37°C for 30 min using T4 ligase according to the pCas9 protocol (Huang et al., 2016). The strategy of multiple editing at cre sites used in this study is depicted in Figure 2.

E. coli Top10 cells were used to construct and amplify plasmids. Ligations of gRNA-annealed oligos and BsaI-digested pCas9 were gently mixed with competent E. coli Top10 cells at a ratio of 1:200 and incubated on ice for 30 min, then heated at 42°C in a water bath for 90 s, and chilled for 1–2 min on ice. The cells were then incubated in preheated LB medium (150 μl) at 37°C for 45 min, then plated onto 25 μg/ml chloramphenicol selective media and incubated overnight at 37°C. The following morning, transformants were picked and incubated in LB medium. After cell growth, plasmids were extracted using an appropriate kit and the pCas9-Cac, pCas9-Cpt and pCas9-Cax constructs were verified by DNA sequencing.

Protoplasts of B. pumilus wild-type (WT) were transformed with plasmid DNA using a modified method described by Chang and Cohen (1979). A single colony of B. pumilus WT was inoculated into 5 ml of LB medium and cultured overnight at 37°C in a shaker (200 rpm). The overnight culture was diluted 50-fold into 5 ml of fresh LB medium and grown for 8 h at 37°C (200 rpm). Cells were harvested by centrifugation (9,000 × g, 10 min, 4°C). The supernatant was removed and cells were resuspended in 1 ml of SMMP solution (equal volumes of 2× SMM buffer and 4× Penassay broth). Lysozyme powder was added to 0.4 mg/ml, mixed, and incubated at 37°C for 45 min (100 rpm) to prepare protoplasts (globular appearance). Protoplasts were harvested by centrifugation (5,000 × g, 10 min, RT), the supernatant was poured off, and the cell pellet was resuspended gently in 0.05 ml SMMP. Plasmid DNA (0.2 μg) was mixed with 0.05 ml of protoplasts and added to a microtube containing 150 μl of polyethylene glycol (PEG) 6,000. After gentle mixing, the sample was incubated at 37°C for 2 min in a water bath shaker to allow plasmids to enter protoplasts, and 0.5 ml of SMMP was added for termination. Protoplasts were harvested by centrifuging (10,000 × g, 7 min) and the supernatant was discarded. SMMP (300 μl) was added and incubated for 60–90 min at 37°C in a water bath shaker (100 rpm) to allow for expression of the antibiotic resistance marker. Because there is no B. pumilus replication site in pCas9, the pCas9-Cac, pCas9-Cpt and pCas9-Cax plasmids were integrated into the genome. Cells were plated on CMR-selective regeneration medium containing chloramphenicol to select transformants. The three plasmids were successively transformed into B. pumilus WT three times. We engineered the B. pumilus LG3145 strain from eight transformants through DNA sequencing.

To analyse the effects of multiple cre sites editing on phenotype, WT and mutant strains were grown in high-glucose fermentation medium (GYN; 10–30% glucose, 2% yeast extract, 0.05% MgSO₄•7H₂O, 1% urea) in triplicate at 37°C with shaking at 220 rpm. A 1 ml sample of culture was diluted 3-fold every 2 h and the optical density at 600 nm (OD₆₀₀) was measured using a T6 UV spectrophotometer using a Purkinje General Instrument (Beijing, China) to plot the growth curve. Cells were harvested by centrifugation at 10,000 × g after culturing for 12 h, washed three times with sterile water, and diluted 100-fold for morphological assay. A 3 μl sample of cell diluent was placed in the centre of a 13 mm × 18 mm piece of mica, freeze-dried for 8 h, and observed by scanning electron microscopy (SEM) using a Zeiss Merlin Compact instrument (Jena, Thuringia, Germany). Cells for atomic force microscopy (AFM) analysis must be washed repeatedly and diluted at appropriate times. A sample of 3–5 μl was placed on a 25 mm × 25 mm fresh, sterile mica sheet and freeze-dried for 8 h before analysis by a Bruker ICON instrument (Berlin, Germany).

We selected the ScanAsyst-air scanning mode to image cells of each strain. For the probe, a silicon tip on a nitride lever was employed, with a cantilever length of 115 μm and a light spring constant of 0.4 N/m. The morphology of colonies on LB agar plates was observed by stereo microscopy using a Motic str6 instrument (Xiamen, China) with a silicon tip on a nitride lever.

The supernatants of LG3145 and WT strains were collected from LB medium cultures by centrifugation (6,000 × g, 20 min, 4°C) and mixed with precooled 12.5% trichloroacetic acid. Following freezing (−20°C for 12 h), the supernatant was centrifuged at 4°C for 5 min to collect proteins. The precipitate was washed 3–5 times with 80% precooled acetone to remove residual trichloroacetic acid and stood for 30 min to completely evaporate acetone. Some precipitates were dialysed for 5 h and freeze-dried for 8 h, then observed using an Oxford X-max20 energy dispersive spectrometer (EDS) and by SEM; other precipitates were dissolved in 1 ml ddH₂O and mixed with 5 ml Folin-phenol reagent A and 0.5 ml reagent B, both purchased from Lanji technology (Shanghai, China). After standing at ambient temperature for 30 min, the mixture was analyzed using a spectrophotometer at 700 nm (Wan et al., 2009; Han et al., 2019). The protein concentration was calculated using the equation y = 0.2823x + 0.1011 (y-axis = absorbance, x-axis = protein concentration).

For pigment analysis, minimal medium (MM; 3.48 g KH₂PO₄, 1.5 g Na₂HPO₄.12H₂O, 3.96 g (NH₄)₂SO₄, 0.7 g MgSO₄.7H₂O, 0.01 g yeast extract, 1 L ddH²O, pH 7–7.5) supplemented with 2.5% (w/v) glycerol or glucose was used to culture each strain at 35°C with shaking at 220 rpm. The absorbance of extracellular pigments was measured from 200 to 800 nm by a Shimadzu UV-3600plus spectrophotometer (Kyoto, Japan). All experiments were repeated in triplicate.

The cre sites of seven target genes were edited by transforming UgRNA:Cas9 expression plasmids pCas9-Cax, pCas9-Cac and pCas9-Cpt into WT strain protoplast following the strategy shown in Figure 2. Chloromycetin (25 μg/ml) was added to the CMR media used for selecting transformants harbouring the UgRNA:Cas9 system.

The promoter regions, 500 bp upstream fragments of target genes ackA, ntdA, ptsH, budA, acsA, xylA and sucC, were amplified by PCR (Supplementary Figure S1) and mutations of the seven target genes were confirmed by DNA sequencing (Supplementary Figure S2). Surprisingly, there were eight mutants with the same genotype among the 10 transformants (success rate = 80%).

The mutation maps of each gene (top of Supplementary Figure S2) show that all point mutations occurred around the target fragments, and the closer the mutations to the centre, the denser they are, and most are distributed along the direction from PAM to target fragment. The number of mutations was significantly positively correlated with the identity of UgRNAs shown in Table 1 (except for xylA). Most of the mutations are transitions, with almost no deletions. Thus, a UgRNA:Cas9 system with more than 50% identity can edit multiple genes at the same time. In fact, the genomes of the mutants were determined by next-generation sequencing (NGS) and the results indicate that most of the mutation sites occurred around the seven target fragments, but some mutation sites occurred at other sites. The occurrence of off-target sites editing appeared to correlate with the similarities of UgRNAs. These relationships will be investigated in future work.

According to editing characteristics, the DNA strand around the target may be twisted and rolled, especially the DNA strand in the binding direction of UgRNAs, and the higher the similarity between UgRNA and target fragment, the larger the curl range. Because the UgRNAs were partially complementary to the target fragments, Cas9 endonuclease seems to wobble within this region and recognise multiple PAM sites for cleaving, which can be verified by the locations and the corresponding PAMs of mutation points shown in Table 2. The results indicate that all mutations occurred in the promoters, which include 1–2 bp mutations at target fragments according to the preset PAM, especially the edited sites of ptsH, acsA and sucC which coincided exactly with cre sites. There were many off-target mutation points with an efficient PAM nearby, such as the transversion T > A(A > T) near the -10 domain of ackA, with a CAG after10 nt; the transversions T > A and A > T of budA, also with TGG and AGG after two and five nucleotides respectively; the point mutation T > G(A > C) of ntdA with a CAG after 2 nt; and the mutation C > T before cre of xylA with an efficient PAM ‘TAG’. All data are marked with a grey shadow in Table 2. However, there were a few distinct off-target mutations in sucC, and only one base exactly edited at the target fragment, possibly because the lower the similarity of UgRNAs, the fewer the number of DNA fragments involved. This phenomenon further proves that UgRNA:Cas9 may cause swing editing of Cas9, and the swing range is affected by UgRNA identity.

The DNA sequencing analysis described above showed that multiple cre sites of B. pumilus LG3145 were edited successfully. Additionally, various phenotypes of LG3145 were markedly altered, including resistance to sugar, capsules outside cells, and white flocculent protein floating on the medium upper layer. We subsequently analysed changes in carbon catabolism, and explored the relationships between mutations and phenotypes.