Generation of chromosome-based recombinase and Cas9 inducible expression E. coli strains
As the first step to develop a novel plasmid mutation method, we engineered chromosome-based recombinase and Cas9 inducible expression strain E. coli LS4515 by knocking the L-rhamnose -inducible cas9 expression cassette into the aacC1 locus of the Red operon integrative strain E. coli LS-GR (Song et al. 2010). We also cloned a cas9 variant encoding Cas9 K848A, K1003A, and R1060A (hereafter refered as Cas9-mKKR) and knocked it into the same aacC1 locus, resulting in E. coli LS4516. Cas9-mKKR was reported to improve the Cas9 gene editing efficiency by decreasing the nonspecific DNA interaction (Halperin et al. 2018, Slaymaker et al. 2015). The recombineering-mediated cassette knock-in scheme and genotype analysis of E. coli LS4516 are shown in Supplementary Figure 1. Chromosome-based tightly regulated Red operon and Cas9 expression strain obviates red operon and cas9 plamid transformation and elimination steps, which is especailly especially for plasmid modication as only mutated plasmid would exist in the cell.
The recombineering function and mutational efficiency of E. coli LS4515 and E. coli LS4516 were subsequently assessed with lagging strand oligo RC407 and leading strand oligo RC408 via bla (ampicillin resistance gene) opal mutation repair. The mutagenic oligo was designed such that besides the target nucleotide mutation (T to C), the first four bases at the 5'-end are phosphorothioated and nucleotide mutations were also introduced at the wobble positions of the two amino acids flanking the target amino acid (Sawitzke et al. 2011). Repair of the opal mutation of the bla would restore the cell’s ability of growth on LB plate with Ap, thus the recombinant clones could be easily differentiated and the recombineering efficiency could be compared by the number of ApR clones.
After co-electroporation of pLS4544 (a pET-based plasmid cloned with the opal mutation of bla) with either RC407 or RC408 into recombinase-proficient electrocompetent cells of E.coli LS4515 and E. coli LS4516, 100 randomly picked clones from each transformation were duplicated on LB agar plate with or without Ap. It showed that for E. coli LS4515, of three transformations, an average of 4.7% and 2.0% ApR clones were obtained for RC407 and RC408, respectively, for E. coli LS4516, of three transformations, an average of 5.3% and 2.3% ApR clones were obtained for RC407 and RC408, respectively. Each 10 clones from RC407 and RC408 oligo transformations were sequenced and bla opal mutations were all successfully repaired. It thus concluded that E.coli LS4515 and E. coli LS4516 exhibited roughly the same recombineering efficiency, this result was not unexpected as the only difference between the two strain is the cas9 region. Consistent with previously reported (Ellis et al. 2001), the lagging strang oligo was more recombinogenic than that of the leading strand oligo.
Establishment of CRM via bla repair
Having demonstrated that the recombineering functions of E.coli LS4515 and E. coli LS4516 and strand preference in recombineering, we tried to establish CRM by adding sgRNA expression plasmid to the electroporation.
The rationae of CRM and CRSM is that after co-electroporation of target plasmid, mutagenic oligo (ssDNA) and sgRNA expression plasmid into the recombinase and Cas9-proficient cells, recombinase-catalyzed homologous recombination between mutagenic oligo and its allele in the plasmid enables the DNA mutation in a locus-specific manner. The mutation sequence was designed either in the protospacer-adjacent motif (PAM) sequence or in the seed sequence (the first 12 nt of the 20 nt spacer), both are key regions for programmed Cas9 cleavage. As the recombineering efficiency is usually 10–5~10–4, more than 99% of the plasmids are not mutated. The targte locus of the none-mutated plasmid would be recognized and cleaved by RNA-guided Cas9, causing DNA break and disappear, therefore only the transformant clones harbor mutated plasmid would survive. By this way, CRISPR/Cas9-assisted ssDNA recombineering greatly increases the recombineering efficiency and facilitates plasmid variant selection. In principle, Cas9 acts as a counterselection marker by cleaving the unedited DNAs.
CRM of bla repair was performed by co-electropoporation of target plasmid pLS4544, bla sgRNA expression plasmid pLS4569 and lagging strand oligo RC407 into the recombinase and Cas9-proficient electrocompetent cells of E.coli LS4515 or E. coli LS4516, 100 randomly picked clones from each transformation were duplicated on LB agar plate with and without Ap. An average of 62% ApR and 82% ApR clones were obtained for E.coli LS4515 and E. coli LS4516 from three transformations, suggesting that the ssDNA recombineering efficiency was significantly improved via Cas9-assisted counterselection and Cas9-mKKR was better than wild-type Cas9 for CRM. The sequencing results of bla mutation and CRM-mediated bla repair are shown in Fig. 2A.
(Fig. 2 inserted here)
CRM of NanA E192N and NanA S208G
We next tested CRM of NanA without phenotypic selection. NanA (E. coli N-acetyl‑D‑neuraminic acid aldolase) catalyzes the condensation between N-acetyl-D-mannosamine and pyvuvate for the chemenzymatic synthesis of N-acetyl-D-neuraminic acid (Neu5Ac), a crucial intermediate for the chemical synthesis of antiviral agent Zanamivir (von Itzstein 2007). E192 and S208 are within the active sites of NanA, E192N and S208G mutants obtained via OE-PCR were used for the chemenzymatic synthesis of Neu5Ac mimetics (Campeotto et al. 2010).
As E. coli LS4516 harbors a chromosomal nanA gene, to avoid mistargeting, we deleted the nanA gene via combination of recombineering and homing endonuclease I-SceI-mediated double strand break (DSB) repair (Gao et al. 2019), resulting in E. coli LS LS4518. Deletion of chromosomal nanA gene and genotype analysis of the deletion mutant are shown in Supplementary Figure 2.
Following the established CRM of bla repair protocol, CRM of E192N and CRM of S208G were performed. For CRM of E192N, we co-transformed nanA gene target plasmid pLS2825 with nanA cloned in pET30a(+) (Gao et al. 2019), E192N sgRNA expression plasmid pLS4580 and mutagenic oligo RC415 into the recombinase and Cas9-proficient electrocompetent cells of E. coli LS4518. For CRM of S208G, we co-transformed pLS2825, S208G sgRNA expression plasmid pLS4582, and mutagenic oligo RC416 into the recombinase and Cas9-proficient cells of E. coli LS4518. Each 10 clones from three repeats were picked for plasmid characterization, it turned out that for CRM of E192N, 10/10, 10/10, and 9/10 correct mutations were observed, for CRM of S208G, 10/10, 8/10, and 10/10 correct mutations were observed. Representative CRM of E192N and CRM of S208G sequencing results are shown in Fig. 2B and Fig. 2C, respectively.
CRM of both E192N and S208G
The nearly 100% CRM efficiency of E192N and S208G promoted us to test whether E192N and S208G double mutation could be obtained simultaneously by co-electroporating target plasmid pLS2825, sgRNA expression plasmids pLS4580 and pLS4582, each 2.5 mg of mutagenic oligo RC415 and RC416 into the recombinase and Cas9-proficient electrocompetent cells of E. coli LS4518. In the first transformation, of 20 clones tested, we got one wild-type, three E192N, and 16 S208G clones. The unbalanced ratio urged us to use half amount of the DNAs used in CRM of S208G (25 ng of pLS4582 and 1.25 mg of RC416) in the next transformation. Of 20 clones tested, we got one wild-type, four E192N, 16 S208G, and one double mutation clones (Fig. 2D). Note that the first nucleotide in S208G part of the double mutation clone was not mutated (which should be t instead of c), yet as ATC triplet nucleotides (red arrow below the sequencing triplet in Fig. 2D) encode the same isoleucine as that of ATT, we didnot pursue the matter further. The persisted, unbalanced mutation ratio might due to the different Cas9 cleavge efficiency between the RNA-Cas9 complex of E192N and S208G or different homologous recombination efficiency of the mutagenic oligos for E192N and S208G or both. Further, it might be possible to increase the ratio of double mutation by adjusting the amount of plasmid or oligo or use of the plasmid cloned with two sgRNA expression elements, or the combination of them.
CRSM of NanA E192 and CRSM of NanA S208
Finally, we tested saturation mutagenesis of NanA E192 and NanA S208. CRM was modified to CRSM by including degenerative nucleotides NNK (N = A/G/C/T, K = G/T) in the mutagenic oligo. We co-electroporated target plasmid pLS2825, sgRNA expression plasmid pLS4580 and mutagenic oligo RC417 for CRSM of NanA E192, or sgRNA expression plasmid pLS4582 and mutagenic oligo RC418 for CRSM of NanA S208. For CRSM of NanA E192, a total of 126 clones from one transformation were randomly picked for plasmid extraction and sequencing, it turned out that all clones were mutated (mutation efficiency was 100%). For CRSM of NanA S208, a total of 103 clones from one transformation were randomly picked for plasmid extraction and sequencing, it turned out that all but one clone were mutated (mutation efficiency was 99.1%). For both mutations, all 21 variants (including one stop codon variant) were successfully generated. Sequencing results of each representative variant for CRSM of NanA E192 and the CRSM of NanA S208 are shown in Fig. 3 and Fig. 4, respectively. The statistics of the sequencing results of CRSM of NanA E192 and CRSM of NanA S208 are summarized in Table 1 and Table 2, respectively.
(Fig. 3 and Fig. 4 are insterted here)
(Table 1 and Table 2 are insterted here)
As seen from the tables, the total number of purines and the total number of pyrimidines for the first two positions of NNK are roughly 50% for both CRSMs, and while the third position of E192 showed unbalance proportion (80 G:46 T), the third position of S208 is nearly identical (53 G:50 T), implying that no obvious nucleotide bias exists for CRSM.
Recently, methods involved in vitro Cas9 (She et al. 2018) or FnCas12a (Dong et al. 2020) cleavage, T5 exonuclease digestion, Klenow fill-in, T4 ligase ligation using annealed dsDNA or single-strand oligo, and transformation were reported for CRIPSR-mediated plasmid mutagenesis, however, the proesses are more complicated and more expensive than the method reported herein. Zhang et al (2020) reported (no experimental procedure was provided) the result of ssDNA recombineering and CRISPR/Cas9-mediated bacterial artificial chromosome (BAC) mutagenesis, perhaps due to the large size of BAC, one out of four clones was correct recombinant.
In this study, we extended the bacterial genome editing (Garst et al. 2017, Oh and van Pijkeren, 2013) to plasmid-based gene mutation. CRM and CRSM circumvent the tedious cloning steps, and up to 100% efficiency were obtained for both. One outstanding advantage of CRSM is that all twenty variants could be obtained in just one transformation and screening of the saturation mutagenesis variants could be finished in one day. By contrast, OE-PCR-mediated mutation might not be able to get all the variants in one round of mutation. For example, one round of OE-PCR-mediated saturation mutategensis of NanA E192 resulted in 15 codons, and single OE-PCR was needed to get the other five variants in the next round (Campeotto et al. 2010). Condition optimization, including adjustment of plasmid and oligo amount, rational spacer design or utilization of other gene editing nucleases, may further reduce the number of clones covering all the variants. Alternatively, degenerate codons NDT or VHG may replace NNK to reduce the number of redundant clones. Conclusively, we hope that CRM and CRSM strategies might accelerate high-thoughput gene mutation studies.