At first, we employed the pCas system, which consists of both the cas9 gene and a λ-Red recombineering system (Jiang et al., 2015). As pCas contains the temperature-sensitive replicon, repA101(Ts), the system needs to be operated at 30°C, which is time-consuming. We isolated a RepA mutant (RepA_A56V), namely pCasM (Figures 2A,B), which both supported the culture of bacteria at 37°C and facilitated the rapid elimination of RepA_A56V plasmids in the absence of antibiotics (Figure 2C). This mutant was therefore employed for genome editing in this study and may also be used in other applications that require final elimination of the plasmids.

If the genomic sites to be edited are within the PAM or protospacer sequence, they can easily be manipulated because of the disruption of Cas9 recognition. However, if the sites to be edited are not within a protospacer or PAM sequence, conventional techniques cannot achieve clean editing because of PAM or protospacer modification. Hence, the two-step strategy employed here is superior in that it avoids the restriction of PAM availability. We took the editing of the lacZ gene in E. coli MG1655 as an example. At first, pCasM was transformed into MG1655 to allow for the induction of expression of the λ-Red recombineering system, and with its help, the ampicillin resistance gene (bla) was successfully inserted into lacZ by HDR. As a result of ampicillin selection, all transformants were positive colonies (Supplementary Figure S2A). Then, an sgRNA (sgRNA_bla) was designed to target the bla gene (scheme shown in Supplementary Figure S1) and the sgRNA plasmid was transformed into MG1655ΔlacZ::bla harboring pCasM. Jiang et al. (2015) acquired high editing efficiency by combining gRNA-expressing plasmid and donor DNA into a plasmid-borne editing template. Based on this, we attempted to co-transform the optimized two-plasmid CRISPR/Cas9 system along with linear donor dsDNA with the aim of achieving effective genome editing. When the donor dsDNA for HDR was absent, virtually no transformants were obtained (Supplementary Figure S2B), demonstrating that sgRNA_bla could be used to guide Cas9 to efficiently cleave the bla gene.

For the next step, co-transformation of sgRNA_bla and specific linear donor dsDNAs was conducted to accomplish genome editing, and in total three different donor dsDNAs were provided: 1) containing the ochre mutation in the lacZ initiation codon, 2) containing an inserted 6 × His encoding DNA sequence (CDS) after the ‘ATG’ initiation codon of lacZ, and 3) containing a fusion of the amilGFP gene to the 5’-end of the lacZ CDS (Supplementary Figure S2B). With the employment of the λ-Red recombineering system, HDR efficiency was largely increased. For each experiment, 20 colonies were randomly picked for colony PCR to confirm the genome editing efficiency, and high efficiency was found for all three editing experiments (Figure 1B and Supplementary Figure S2B). Sanger DNA sequencing further confirmed that all PCR-positive colonies contained the correct target DNA sequences (Figure 1B and Supplementary Figure S3).

Besides of sgRNA_bla, another two sgRNAs were designed for specifically targeting ARC (i.e., sgRNA_bla1 and sgRNA_bla2), both of which showed high editing efficiency (Supplementary Figure S4), demonstrating the robust of this strategy. In addition, different lengths of homologous arms of the donor dsDNAs were also tested. We noticed that a large number of colonies failed to be amplified with colony PCR with primers surrounding the target lacZ site, suggesting there were sequence rearrangements around the site, which was similar to the observation previously reported (Qi et al., 2013). Although the rearrangement ratio was found to be in a negative relationship with the length of the homologous arms, the positive ratio of the rest colonies that could be successfully PCR amplified was unaffected (Supplementary Figure S5).

Moreover, the engineered mutants were verified by phenotypic tests, which included blue/white colony screening, a western blot assay and fluorescence microscopy. As shown in Figure 3, the mutant containing the ochre mutation failed to turn blue when grown on LB plates supplemented with isopropyl β-D-1-thiogalactopyranoside (IPTG) and 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-gal). For the mutant containing a His-tag on the N-terminal of LacZ, the western blot result clearly showed expression of the His-tagged LacZ. Similarly, expression of the AmilGFP-LacZ fusion protein could be observed by a fluorescence microscope in the amilGFP-lacZ mutant. Considering the expression of fluorescence by the mutant, we also employed FACS to precisely measure the recombination efficiency during the insertion of amliGFP, and the results showed that more than 93% edited cells successfully expressed AmilGFP (Supplementary Figure S6), which was consistent with the results of Figure 1B. In all mutants, the plasmid system was easily eliminated by culturing overnight with shaking in the absence of antibiotics (Supplementary Figure S7). In short, optimization of the CRISPR/Cas9 and two-step strategy, as described here, allowed for precise and efficient manipulation of the E. coli genome.

To edit the 3′-end of an essential gene, similar procedures can be employed as mentioned above, where an ARC can be inserted nearby the stop codon but outside of the CDS. Whereas, to edit the 5′-end or the internal region of an essential gene, with or without an independent promoter, a modified version of this method can be used, where a synthetic ribosome binding site (BBa_J61101) is placed after the ARC, which thus allows for the normal transcription and translation of the essential gene (Figure 4A). To test this system, the frr gene, which is an essential gene encoding a ribosome releasing factor (Janosi et al., 1994), was selected, and according to the Sanger sequencing results, the His-tag DNA sequence was successfully inserted after the translation initiation codon in the frr gene (Figure 4B). To the best of our knowledge, this is the first report describing the genome editing of essential genes using CRISPR/Cas9 in E. coli.

It is worth mentioning that the ARC is employed in our method as both a positive selection marker during the first step and a negative selection marker (the cleavage target of Cas9) during the second step. Although the Cas9-assisted dsDNA break may increase the editing efficiency of the first step, it is not necessary when the λ-Red recombineering system is employed. Therefore, as shown in Supplementary Figure S8, our method provides a simplified experimental procedure and is time saving.

Recently, the Cas9-mediated two-step methods have been reported that allow for genome editing (including point mutations). Specifically, during the first step, artificial protospacer sequences are introduced to replace the original protospacer near the target sites using the CRISPR/Cas9 system. Then, through employing a second sgRNA that recognizes the introduced artificial protospacer sequence, the target sites will be mutated, and meanwhile the artificial PAM or protospacer sequence will be back-mutated to its original sequence, generating a clean mutant with only the target sites mutated (Biot-Pelletier and Martin, 2016). Wang et al. (2016) reported a similar approach. The PAM or protospacer used in the first step was modified or deleted, whereas the artificial PAM was created at the target site. Then in a second step, through CRISPR/Cas9 editing, the modified or deleted PAM or protospacer sequence was changed back into its original sequence and the artificial PAM was changed into the desired mutation (Wang et al., 2016). Although these methods are effective, their first steps require a proper PAM site near to the target sites, which may limit their application. That is, these methods cannot edit target sequences without a PAM sequence in the vicinity. Moreover, to avoid continuous cutting of the modified genome, the sgRNA-expressing plasmid must be cured before the second step of genome manipulation, which could be time consuming. By contrast, our strategy (Figure 5) allows for convenient and prompt curing of the sgRNA-expressing plasmid, saving much time.

In summary, we describe an efficient two-step genome editing approach that integrates both the CRISPR system and an ARC. A specific modification of the target genome site can be generated with ease by supplying a template DNA containing the desired mutation. As the method is versatile and does not require any PAM sites in the genome, it can therefore be used to conveniently edit any DNA sequences.

Escherichia coli strain DH10B was used for plasmid cloning and strain MG1655 was used for genome editing analysis. Plasmids pTarget and pCas were kindly provided by Prof. Sheng Yang (Jiang et al., 2015). Unless indicated, all E. coli strains were grown in Luria–Bertani (LB) medium and incubated at 37 °C. All strains and plasmids used in this study are listed in Table 1 and all primers are listed in Table 2.

First, we adopted the primary pCas system, as pCas requires a temperature of 30°C. However, to optimize the efficiency of the system, we attempted to culture bacteria at 37°C with several inoculations and a RepA mutant was obtained, which was designated pCasM.

The optimized plasmid pCasM was transformed into E. coli MG1655 and a transformant was used to prepare electrocompetent cells. The expression of the λ-Red recombinases was induced by addition of 10 mM arabinose. When the OD₆₀₀ reached 0.6, cells were pre-chilled on ice for about 10 min before being washed three times with 10% cold glycerol. About 5 ml of cells were finally concentrated to 50 μl for each electroporation reaction.

Both upstream and downstream homologous arms were PCR amplified with primers listed in Table 2, and were then fused with the ampicillin resistance cassette (the ARC used in this study) by fusion PCR. About 50 μl of electrocompetent cells were mixed with 300 ng of the ARC, and electroporation was performed in a 2-mm cuvette using the GenePulser Xcell^TM (Bio-Rad, USA; 2.5 kV, 200 Ω). Transformants were suspended immediately in 1 ml of ice-cooled LB liquid, and recovered by shaking at 37°C for 1 h before being spread onto LB agar medium containing ampicillin (100 μg/ml). After being incubated overnight at 37°C, transformants were verified by colony PCR and DNA sequencing, using the primers listed in Table 2.

To construct psgRNA_bla, a pair of primers were used for PCR amplification using pTarget (Jiang et al., 2015) as the template. A 20-bp spacer sequence (N20) was specifically designed to target the ARC, and this was incorporated into the primers. The amplicon was digested with DpnI (NEB, USA) to remove the template, and was then transformed into DH10B to obtain the desired psgRNA plasmid, namely psgRNA_bla, followed by verification by Sanger DNA sequencing.

Donor dsDNA fragments usually contained a 300–500 bp homologous arm on each side. For mutation of the lacZ initiation codon and labeling with an N-terminal His-tag, DNA sequences of “TAA” and “CACCACCACCACCAC” were directly incorporated into primers to amplify the upstream and downstream homologous arms. For the fusion of amilGFP to lacZ, two homologous arms and the amilGFP CDS sequence were separately amplified and then assembled together by fusion PCR. For labeling of the frr gene with the N-terminal His-tag, the DNA sequence of “CACCATCATCACCACCAT” was directly incorporated into the primer for amplification of the downstream homologous arm. Both upstream and downstream homologous arms were then assembled by fusion PCR. All PCR products were purified by gel electrophoresis prior to electroporation.

Cells harboring the ARC and pCasM were used for the preparation of electrocompetent cells, following the same procedure as described above. The electroporation parameters were the same as described, and 200 ng psgRNA plasmid and 300 ng donor dsDNA were added to each electroporation reaction. Cells were recovered at 37°C for 1 h before being plated on LB agar containing kanamycin (50 μg/ml) and spectinomycin (300 μg/ml) and incubated overnight at 37°C. Transformants were verified by colony PCR and subsequent Sanger DNA sequencing.

To eliminate psgRNA and pCasM plasmid, a positive colony was selected and inoculated into 5 ml liquid LB medium with no antibiotics for overnight culture. Then, cells were streaked onto LB agar to allow for the formation of single colonies, which were then checked by a growth test on plates containing kanamycin (50 μg/ml), spectinomycin (200 μg/ml) or no antibiotics. When a colony was sensitive to both antibiotics, it would have lost both plasmids and could therefore be used for further analysis. For curing of the psgRNA plasmid, an alternative method was to grow the cells harboring psgRNA in liquid LB medium with IPTG (0.2 mM) to induce the expression of an sgRNA specifically targeting the psgRNA plasmid, which would then lead to the cleavage of psgRNA by Cas9 expressing from pCasM plasmid.