Efficient dual-negative selection for bacterial genome editing

Gene editing is key for elucidating gene function. Traditional methods, such as consecutive single-crossovers, have been widely used to modify bacterial genomes. However, cumbersome cloning and limited efficiency of negative selection often make this method slower than other methods such as recombineering. Here, we established a time-effective variant of consecutive single-crossovers. This method exploits rapid plasmid construction using Gibson assembly, a convenient E. coli donor strain, and efficient dual-negative selection for improved suicide vector resolution. We used this method to generate in-frame deletions, insertions and point mutations in Salmonella enterica with limited hands-on time. Adapted versions enabled efficient gene editing also in Pseudomonas aeruginosa and multi-drug resistant (MDR) Escherichia coli clinical isolates. Our method is time-effective and allows facile manipulation of multiple bacterial species including MDR clinical isolates. We anticipate that this method might be broadly applicable to additional bacterial species, including those for which recombineering has been difficult to implement.


Background
Genetic engineering is fundamental for molecular analysis of genotype-phenotype relationships, and for determining the function of previously uncharacterized genes [1][2][3]. Site-specific mutagenesis can be achieved using different methods. Traditionally, marker-free genetic manipulations were obtained using consecutive single-crossovers mediated by endogenous recombinases [4,5]. A suicide vector is first integrated in the desired location using homologous recombination. Bacteria, in which a subsequent second crossover results in loss of the integrated plasmid, can then be selected using counter-selection markers [6][7][8][9]. However, counter-selection is often suboptimal resulting in a need to screen many clones for the desired event [10,11]. Later, the λ-Red recombineering technology, a phage-based homologous recombination system based on linear DNA transfer and an exogenous recombinase, was introduced [8, [12][13][14][15]. Scarless mutations can be obtained when combining this method with a counter-selection marker [16][17][18][19]. Currently, λ-Red recombineering is the method of choice for introducing genetic manipulations in S. enterica and E. coli [20] but it has been difficult to implement in several other bacterial species such as Pseudomonas aeruginosa. Recently, clustered regularly interspaced short palindromic repeats (CRISPR)-Cas has revolutionized eukaryotic genome editing [21][22][23], but this strategy is more cumbersome for bacteria with limited recombination activities [24][25][26].
Here, we combined several improvements for establishing a time-efficient versatile method for consecutive single cross-overs in multiple bacterial species. We used rapid Gibson assembly of PCR products [27] to generate suicide vectors with dual negative selection mediated by I-SceI and SacB [28,29] (Fig. 1a), which increased counter-selection efficiency to 100% for nearly all tested deletions, insertions and point mutations. We employed an E. coli donor strain that simplifies donor removal after conjugation and avoids common problems with contaminating phages [31]. We used different positive selection markers that enable selection in many bacterial species, including MDR pathogens [32]. Combination of these elements yielded a reliable and fast method for genetic engineering of multiple bacterial species that, in concert with a simplified screening procedure, minimized hands-on time and significantly accelerated genome editing in our lab.

Results
Our goal was a rapid and efficient genetic editing method with minimal hands-on time. For this purpose, we combined rapid plasmid construction using Gibson assembly [27], a phage-free, pir-carrying (for propagation of R6Kγ plasmids), diaminopimelic acid (DAP)-dependent E. coli donor strain JKe201 [31] for plasmid amplification and conjugation, with subsequent facile removal of donor in absence of DAP, and an improved dual-negative counterselection. We generated suicide vectors from PCR fragments with automatically designed primers using Gibson assembly [27]. Each vector carries commonly used genetic elements for conditional propagation ("suicide vector" with pi-dependent replication from R6Kγ), conjugation (oriT, traJ) and selection for two sequential singlecrossovers. For the first positive selection, we used aphA conferring resistance to kanamycin (pFOK, Fig. 1a).
A major limitation to efficient genetic editing using two consecutive single-crossovers has been inefficient counter-selection of the second recombination, in part due to inactivating mutations in the negative selection marker [33]. We tested counter-selection efficiency in multiple Salmonella loci using the commonly used markers sacB or I-sceI (Fig. 1b). sacB codes for levan sucrase, which confers sensitivity to sucrose because of accumulation of the toxic product levan in the periplasm Fig. 1 An optimized method for genome editing in Salmonella enterica. a Map of suicide plasmid pFOK (aphA, aminoglycoside phosphotransferase gene conferring resistance to kanamycin; I-sceI gene encoding meganuclease; oriT, origin of conjugational transfer; P tetA , tetA promoter; R6K γ ori, pi-dependent origin of replication; sacB, levansucrase gene; tetR, tetracycline repressor gene; traJ, transcriptional activator for conjugational transfer genes; MCR, multi cloning region containing EcoRI and BamHI recognition sites). b Mechanisms of negative selection for SacB and I-SceI, c Efficiency of negative selection for various chromosomal loci (sitABCD deletion -orange, foxA deletion -yellow, ssrB point mutationgreen, and phoQ chimeric insertion -magenta [30]) using either SacB or I-SceI, or a combination of both. Fifty colonies were screened for each mutation. d Schematic representation of the consecutive single crossover procedure. Recombination can occur in one of the two homologous sequences (routes 1 and 2). Only alternate single crossover events involving both homologous sequences lead to the desired mutation, while two consecutive single crossovers in the same regions lead to reversion to wild-type (WT) e Overview of the entire procedure. Ideally, each step can be completed in one working day. f Schematic representation of preferential recombination in the right flanking region. External primers 1 and 2 together with plasmid-specific primers oOPC-614 and oOPC-615 can be used to screen co-integrant clones to reveal such bias and to identify rare variants for promoting mutant generation in the second single crossover. g Recombination bias for foxA deletion. PCR results of ex-conjugant screening using external primer 1 (oOPC-396) / oOPC-614 or external primer 2 (oOPC-397) / oOPC-615. Rare exconjugants (clones 5, 10) with recombination in the non-preferred flanking region were used for subsequent counter-selection [28]. I-sceI codes for the restriction enzyme I-SceI, which causes lethal DNA double-strand breaks when a I-SceI recognition sequence is present on the genome [29]. To assess counter-selection efficiency of SacB or I-SceI singly, we generated plasmid variants (pOPC-001 and pOPC-003) differing just in the counter-selection. Counter-selection was suboptimal for both markers with marker-free clones representing none or only a minority of the recovered colonies (Fig. 1c). Consequently, many colonies had to be tested for finding the desired clones. To overcome this problem, we generated a new suicide vector, pFOK, combining both sacB and I-sceI under the regulatory control of the TetR regulator (Fig. 1a). We tested the TetR system using the green fluorescent protein (GFP) as reporter on the same pSC101 backbone and found no detectable GFP fluorescence above the autofluorescence background in absence of the inducer anhydro-tetracycline indicating limited leakiness in our conditions ( Supplementary Fig. S1). Cells carrying the conditional dual-negative selection cassette under control of the TetR system showed no decrease in cloning efficiency but efficient negative selection in presence of sucrose and anhydro-tetracycline, yielding only, or a large majority, of resolvants that had successfully cured pFOK from their chromosome (Fig. 1c). A similar dual-negative selection has been previously described for Gram-positive bacteria [34]. The ratios for the two alternative resultsmutation or reversion back to wild-typevaried between the individual mutants ( Supplementary Fig. S2).
To expand our gene manipulation method to other bacterial species, including those for which λ-Red recombineering has not yet been established, we used alternative positive selection markers. This included aac (3)-I, coding for a aminoglycoside N-acetyltransferase that confers resistance to gentamicin which can be used as an alternative in bacteria, including Pseudomonas aeruginosa, which are resistant to kanamycin but susceptible to gentamicin (pFOG, Fig. 2a). We confirmed the utility of pFOG by deleting the mexAB operon in P. aeruginosa and observed 50 resolvants among 50 tested colonies (100%) after negative selection. As an alternative, we combined aphA with a second positive marker, tpm, yielding suicide vector pFOKT (Fig. 2b). tpm codes for a thiopurine-S-methyltranferase conferring resistance to tellurite [35]. This plasmid can be used for multi-drug resistant (MDR) bacteria for which the choice of positive selection markers is limited [32]. To limit toxic exposure to volatile dimethyl telluride, we used kanamycin for suicide vector generation and used tellurite only for the positive selection of ex-conjugants. We confirmed the utility of pFOKT by deleting tolC with high efficiency in a multi-drug resistant clinical Escherichia coli isolate [32] and again observed 50 resolvants among 50 tested colonies (100%) after negative selection.
In some cases, the second single-crossover had a high bias for resolution to wild-type loci (instead of the desired mutant). This was usually due to differences in recombination frequency between the two flanking regions. PCR primers (oOPC-614 and oOPC-615) that bind in the plasmid, combined with chromosomal primers outside the flanking regions in the merodiploids (Fig. 1f, g), enabled detection of such biases. For these cases, we selected exconjugants in which the first single-crossover had occurred in the non-preferred flanking region. In these clones, we often observed frequent resolution to mutant loci during the second single-crossover (deletion of foxA, Supplementary Fig. S2).
Altogether, the whole protocol from initial plasmid construction to scar-less sequence-verified mutant strains (Fig. 1e) was completed within five working days with minimal hands-on time for 23 of 30 Salmonella mutants. We obtained all residual mutants after additional optimization of the initial PCR (four mutants), prolonged cultivation times for mutants with reduced growth (two mutants), or screening for biased recombination as shown  Fig. 1f, g (one mutant -ΔfoxA). The Escherichia tolC mutant was also completed within 5 working days, while slow growth of Pseudomonas at 28°C during the negative selection required a total of 6 working days. In all cases, we obtained 100% resolvants after a single round of negative selection confirming the efficiency of our method (the only exception was the ssrB point mutant with 70% resolvants as shown in Fig. 1c).

Discussion
Gene editing enables investigation of gene function. Here, we improved on a widely used method of consecutive single-crossovers. Our newly developed suicide vectors, based on a highly efficient dual negative selection strategy, mitigate the major pitfall of consecutive singlecrossovers: the poor selection of resolvant clones after the second recombination. Thus, our new vectors do not require multiple rounds of counter-selection to ensure resolution of the suicide vector from the recipient strain. One of the two negative selection genes encodes I-SceI which recognizes a specific 18-basepair sequence [36]. While none of the strains used in this study harbored a I-SceI recognition sequence in their genomes, this sequence might be present in other bacteria which would need method adaptation. Gibson assembly enables rapid construction of plasmids with PCR fragments with no need for enzyme digestion and ligation, and no sequence constraints due to restriction sites. Our approach relies on endogenous RecA, but not the heterologous, powerful lambda-red recombinase, which might minimize the risk of secondary mutations. Purifying mutated loci by generalized phage transduction may thus not be required. Our method employs conjugation instead of electroporation (as required for lambda-red methods), which minimizes culture volumes and hands-on time. We anticipate that this method might be broadly applicable to additional bacterial species, including those for which recombineering has been difficult to implement.

Conclusions
Our plasmids and protocols provide facile time-efficient methods for genetic engineering in multiple bacterial species including MDR clinical isolates.

Generation of the suicide vectors
Primers for generating pOPC-001, pOPC-003 and pFOK are reported in Supplementary Table S1. pOPC-001 was obtained by combining the kanamycin resistance cassette and the I-SceI restriction site from pWRG717 [37], the origin of replication (R6Kγ) and origin of transfer (oriT) from pGP704 [6,38] and the tetR and I-sceI locus from pWRG730 [37] using Gibson assembly. pOPC-003 was generated by replacing the tetR and I-sceI locus from pOPC-001 with sacB from pEXG2 [39]. pFOK (5841 bp) was generated by inserting sacB amplified from pOPC-003 downstream of the I-sceI gene on pOPC-001. pFOG (5659 bp) was generated by replacing aphA of pFOK by acc (3)-I. pFOKT (6668 bp) was generated by insertion of tpm [35] between aphA and the multi cloning region (MCR).

Amplification of the upstream and downstream regions
Flanking primers with a 40 bp overlap were designed to amplify 700 bp up-and downstream of the gene of interest using SnapGene® (version 4.0.3) with the Gibson Assembly tool (Supplementary Table S1). Fragments were amplified using a high-fidelity polymerase mix (KOD Hot Start Master Mix, Millipore) and separated on a 1% agarose gel. Vectors were purified from overnight cultures using a plasmid miniprep kit (ZymoPURE™, ZymoResearch). Vectors were digested using EcoRI-HF and BamHI-HF (New England BioLabs) for 1 h at 37°C, or PCR-amplified, and purified on agarose gel. Alternatively, vectors can also be amplified by long-range PCR. Final concentrations of amplificated fragments and digested vectors were measured using a microvolume spectrometer (Colibri®).

Gibson assembly and chemical transformation
Plasmids generated in this study are listed in Supplementary  Table S2. Gibson assembly reaction was performed as described [27]. The reaction mix contained 50 ng of each upand downstream fragments and 150 ng of suicide vector, and Gibson assembly mix 1x (New England BioLabs) in a total volume of 10 μl. The reaction mixture was incubated at 50°C for 20 min. Five microliters of the reaction mixture was added to a 100 μl aliquot of E. coli JKe201 heat-shock competent bacteria and incubated for 30 min on ice. After a heat shock at 42°C for 30 s followed by 2 min on ice, bacteria were resuspended in 1 ml prewarmed SOC medium containing 100 μM of DAP and incubated for 1 h at 37°C. Transformants were selected on Lennox-LB agar plates containing 100 μM DAP (required for growth of JKe201) and either 50 μg/ml kanamycin or 15 μg/ml gentamicin. Clones were screened using PCR with primers oOPC-614 and oOPC-615 (Supplementary Table S1).

Conjugation and selection of the first homologous recombination event
The recipient S. Typhimurium and E. coli strains were inoculated in 2 ml of LB containing no antibiotics at 37°C. P. aeruginosa was inoculated in 2 ml LB without antibiotics at 42°C. The donor E. coli strain was inoculated in 2 ml of LB containing 100 μM DAP but no antibiotics. Five hundred microliters each of overnight cultures of the donor E. coli strain and the recipient strain were washed with LB, mixed and centrifuged together. The pellet was resuspended in 50 μl LB and deposited on 22 mm filter membranes with 0.45 μm pores (Millipore, Merck) on a pre-dried LB agar plate. After mating for 6 h at 37°C, bacteria were scraped from the filter membrane and resuspended in 1 ml LB. Merodiploid S. Typhimurium (pFOK) were selected on LB plates containing 90 μg/ml streptomycin and 50 μg/ml kanamycin at 37°C for at least 16 h. E. coli (pFOKT) and P. aeruginosa (pFOG) merodiploids were selected on LB plates containing 10 μg/ml tellurite or 30 μg/ml gentamicin, respectively. Clones grew on tellurite to form black colonies.

Counter-selection of the second homologous recombination event
At least three trans-conjugant colonies were combined and grown for 4 h at 37°C in 2 ml of LB. Bacteria were then streaked on freshly prepared LB-no salt agar plates [24] containing 20% sucrose and 0.5 μg/ml AHT. Plates were incubated at 28°C protected from light for at least 24 h. Colonies were screened for the desired mutation using PCR with external primers (Supplementary Table S1). Mutants were confirmed by DNA-sequencing (Microsynth.ch).