Versatile allelic replacement and self-excising integrative vectors for plasmid genome mutation and complementation

ABSTRACT The ability to better understand the function of proteins expressed by bacteria has typically relied upon the development of genetic mutant strains. This approach has been especially challenging for plasmid-encoded genes, as most of the previously described allelic replacement vectors are inefficient for plasmid genome mutation as they either rely on plasmid-derived counterselection toxins or depend on other strategies such sacB, tetA, and rpsL which have been proven to be less efficient for mutant selection. Integrative vectors lack chromophore indicators, thus requiring laborious screening or excision of the vector’s backbone relies on the introduction of a flippase (FLP)-expressing plasmid. The allelic replacement vector, designated here as pDG1, expresses an X-Gal hydrolyzing enzyme (BgaB) that can be used for blue/white screening allowing identification of colonies that integrated and successfully removed the mutagenesis plasmid without a bias for those still carrying it. pDG1 was further improved by including the rhamnose-inducible Tse2 toxin as a potent counterselection system. The efficacy of pDG1 was validated by deleting portions of the plasmid-encoded VirB4/D4 type IV secretion system and aerobactin-synthesizing operons in Salmonella enterica. The integrative vectors, which contain an ΦC31attP site and genes encoding ΦC31 integrase (int), can seamlessly integrate to target ΦC31attB on Salmonella plasmids or chromosome. These vectors were improved by inserting bgaB and FLP-encoding genes so that, following integration, most of the vector’s backbone encompassing int, bgaB, and FLP genes can be excised by FLP, without the need for another FLP-expressing vector, creating white colonies carrying a stably integrated target gene. As such, we were able to integrate a 9.3-kb DNA fragment to Salmonella chromosome and flipped out most of the integrated vector in one step, leaving the target fragment in the chromosome. IMPORTANCE In spite of the dissemination of multidrug-resistant plasmids among Gram-negative pathogens, including those carrying virulence genes, vector tools for studying plasmid-born genes are lacking. The allelic replacement vectors can be used to generate plasmid or chromosomal mutations including markless point mutations. This is the first report describing a self-excising integrative vector that can be used as a stable single-copy complementing tool to study medically important pathogens including in vivo studies without the need for antibiotic selection. Overall, our newly developed vectors can be applied for the assessment of the function of plasmid-encoded genes by specifically creating mutations, moving large operons between plasmids and to/from the chromosome, and complementing phenotypes associated with gene mutation. Furthermore, the vectors express chromophores for the detection of target gene modification or colony isolation, avoiding time-consuming screening procedures.

S almonella enterica is a leading cause of foodborne illness in the United States (1)   and around the globe (2).This bacterium is often known to harbor multidrug resistance plasmids of different incompatibility (Inc) groups (3)(4)(5)(6).Some of the plasmids in Salmonella are known to carry well-studied virulence factors such the spv operon (7,8), less studied virulence factors such as VirB4/D4 type IV secretory systems (VirB/D4 T4SS) (9), and iron acquisition systems including aerobactin and the Sit iron transport system (5).A challenge for studying these and other plasmid encoded systems is the lack of efficient vector tools for modification of the plasmid genome in Salmonella and related species.
Phage integrases catalyze the recombination between bacterial and phage attach ment sites, often referred to as attB and attP, respectively (10,11).Phage integrases have been used for integration of target genes containing attP sites into either a genetically engineered attB site or preexisting attB sites in the genome of Gram-positive (12,13) and Gram-negative pathogens (14,15).With the phage integration system, experiments, including animal studies (13), can be done without the need for antibiotics for plasmid maintenance.This advantage is because of the stability of the integrated vector as excision of the vector, and subsequent loss requires excisionase and other host factors that are typically not encoded by target stain (10,11).Integrative vectors have not been described for Salmonella, underscoring the need for single-copy integrative vectors that can be used for complementation studies and for insertion of large genetic fragments which cannot otherwise be inserted using conventional knock-in vectors.
With regard to gene deletion vectors, excellent tools have been developed for mutation of chromosomal genes in Gram-negative bacterial pathogens (16)(17)(18)(19)(20)(21) including Salmonella (16,18).The lambda red recombinase system has been the most extensively used gene knockout tool in Gram-negative bacteria, including in Salmonella (18,22,23).Although this is an excellent resource for chromosomal gene mutations and in some cases for plasmids (24,25), recombineering has its own limitations related to unintended flippase (FLP)-mediated genetic "scar" formation or off target mutation.CRISPR/Cas9 or I-SceI Cleavage Systems have also been coupled with lambda red recombinase to generate precise chromosomal gene deletion without involving FLP recombinase, which avoids introduction of "scar" following antibiotic resistance gene elimination (20,26).These systems depend on the principle that Cas9 or I-SceI-medi ated double-strand breaks in a chromosome are lethal to the bacterium (19,21).Thus, in principle, cells that have undergone double-strand break repair via homologous recombination with a template provided as a PCR product, plasmid, or oligonucleotides should survive providing an excellent counterselection over cells possessing unrepaired double-strand breaks.Such counterselection may not work for plasmid mutation as double-strand break in plasmid is not lethal.
Another excellent resource for gene mutagenesis is the use of conditionally replicative plasmids, often referred to as suicide vectors, such as those with R6K ϒ replication origin (ori R6K γ ) (15) and plasmids carrying temperature-sensitive origins of replication such as pKO3 (27) and its derivatives (28,29).Plasmids with ori R6K γ can only replicate in the presence of the π protein, allowing target site homologous recombination in strains lacking the pir gene (30,31), forming an event dubbed as single crossing over.Regarding plasmids carrying a temperature-sensitive origin of replication, recombination occurs at the non-permissive temperature (27,32).In these vectors, a second recombination excises the plasmid from the chromosome creating either a wild type or a modified allele in the chromosome.Whatever the case, plasmid loss, and excision events occur very rarely, thus, an appropriate counterselection marker is critical for recovery of desired mutant after allelic replacement.In this regard, sucrose counterselection is an extensively used technique in Gram-negative bacteria (17,27,28,33), but this system confers low counterselection stringency (34)(35)(36).A set of powerful negative counterselection markers expressing inducible plasmid-derived toxins has been shown to provide an excellent stringency in counterselection (36), but vectors express ing these toxins may not be used for plasmid genetic modification as these toxins can be neutralized by their cognate's antitoxins.Thus, the aim of this study was to develop efficient allelic and integrative vectors for plasmid genome studies.The vectors described in this study contain chromophore indicators which allow visual detection of mutants, thereby narrowing to the colonies with target gene deletions or integration and avoiding time-consuming screening procedures.

Bacterial strains and culture conditions
Strains and plasmids used in this study are listed in Table 1.All culture media, antibiotics, and media supplements including sucrose, 5-Bromo-4-chloro-3-indolyl β-D-galactopyra noside (X-Gal), and L-rhamnose were purchased from Fisher Scientific (Pittsburgh, PA, USA).All primers were purchased from Integrated DNA Technologies (Coralville, IA, USA).When needed, antibiotics were added to the media at a final concentration of 20 µg/mL for zeocin, 75 µg/mL for hygromycin B, and 30 µg/mL for nourseothricin and chloram phenicol.Culture media were supplemented with 0.2% L-rhamnose and 30 mg/mL X-Gal as needed.All restriction enzymes, T4 DNA ligase and NEBuilder HiFi DNA Assembly Master Mix (HiFiMix) were purchased from New England Biolabs Inc. (NEB, Ipswich, MA, USA).Strains used in this study were grown in Luria-Bertani (LB) and M9 minimal media as needed.Recombinant plasmids were transformed to Salmonella with a Gene Pulser II (Bio-Rad, Hercules, CA, USA) according to the manufacturer's guideline.

Construction of pKOV-ZeO
Primers used in this study are listed in Table S1.The zeocin resistance marker (zeoR) was amplified from pCRBlunt II-TOPO vector (Invitrogen, Waltham, MA USA) using zeo1 and zeo2 primers and inserted into the NotI-BamHI-restricted pKOV vector (Addgene plasmid #25769, a gift from George Church) in the same orientation as in chloramphenicol resistance gene (cmR).zeo1 was designed by adding the NotI restriction site at the 5′end, and similarly, SacI, XhoI, and BamHI restriction sites were incorporated to the 5′end of zeo2 for cloning purposes.lacZ alpha amplified from V308 pDNR MCS lacZ alpha-Donor (Addgene plasmid #11691, a gift from Tony Pawson) using lac1 and lac2 primers was inserted at the XhoI site, and the final vector was designated as pKOV-ZeO (Fig. S1).Unless otherwise noted, all PCR was performed using a T100 thermalcycler (Bio-Rad) using GoTaqG2 Green Master Mix (Promega, Madison, WI, USA) to verify plasmid inserts/ constructs and mutants.Phusion High-Fidelity DNA Polymerase (ThermoFisher, Waltham, MA USA) was used for amplification of cloning fragments and amplification of rela tively large DNA fragments.PCR conditions were run according to the manufacturers' instructions with annealing temperatures optimized for the primer pairs.

Allelic replacement vector using pKOV-ZeO
To construct an iutA deletion mutant, upstream and downstream flanking regions of the gene were sewed together by overlapping PCR using iuc1, iuc2, iuc3, and iuc4 primers followed by inserting to pKOV-ZeO at the BamHI and SalI sites.The resulting recombi nant plasmid was transformed to Salmonella enterica strain SE163A (5).Cells were grown in zeocin-containing LB media overnight by shaking them at 30°C.This overnight culture was diluted 100-fold in zeocin-supplemented LB followed by incubating with shaking at 42°C for 6 hours.Cultures were then streaked on LB agar supplemented with zeocin followed by overnight incubation at 42°C.A few colonies from the 42°C plate were inoculated to plain LB broth followed by shaking at 30°C overnight.This overnight culture was diluted 10-fold in LB broth and plated on LB agar supplemented with 5% sucrose.Sucrose-resistant colonies were replica plated on plain LB agar and LB agar containing zeocin.Zeocin-susceptible colonies were screened using iuc5 and iuc6 primers to verify iutA deletion, and the resulting mutant was preserved as DGS4 (Table 1).

Construction of pDG1 and pDG2 vectors
To construct the pDG1 vector, we first removed the sacB gene and most of the f1 ori from pKOV-ZeO by inverse PCR using cirF and cirR primers followed by circularizing the PCR product at the NdeI site to create pKOVcir.We then amplified the bgaB β-galactosi dase-encoding gene including its promoter from pMAD (a kind gift of Chia Y. Lee) by the AvrII site containing lac3 and lac4 primers.This PCR product was then cloned at the AvrII site of pKOV-ZeO, replacing the lacZ alpha gene.The resulting vector was designated as pDG1 (Fig. 1).To construct pDG2, we first amplified mCherry from pFP.E227 (Addgene plasmid #138249, a gift from Baojun Wang) using the AvrII site containing mc1 and mc2 primers.The trc promoter sequence was inserted into the 5′ of mc1 for efficient mCherry expression.The PCR product obtained by mc1 and mc2 was cloned to AvrII as above, and the resulting vector was designated as pDG2 (Fig. S2).

Allelic replacement using the pDG vectors
A recombinant plasmid for deleting the iucABCD genes of the aerobactin operon was constructed by sewing upstream and downstream flanking regions of iucABCD using overlapping PCR and iuc7 (contains BamHI restriction site at its 5′end), iuc8, iuc9, and iuc10 primers.The primers were designed in a way that they do not result in frame shifts in downstream open reading frames during allelic replacement.The overlapping PCR product was TOPO cloned in a Zero Blunt TOPO PCR cloning kit (Table 1).The overlapping PCR product was cut from the TOPO vector with BamHI followed by cloning into the corresponding site on the pDG1 vector.The resulting recombinant plasmid was transformed to Salmonella enterica strain SE163A (5) followed by selection on LB plates supplemented with zeocin and X-Gal.The allelic replacement procedure was performed as in pKOV-ZeO, with the exception that the plates were supplemented with X-Gal.Light-blue colonies obtained from the 42°C plates were grown overnight in LB with shaking at 30°C.This overnight culture was diluted 10-fold followed by plating on X-Gal-supplemented LB agar.Colony PCR was performed on white colonies to verify iucABCD operon deletion by using the iuc11 and iuc12 primers, which anneal ∼0.61 kb upstream and ∼0.64 kb downstream of the region flanking the ∼5.6-kbdeletion target.PCR using primers iucbF and iucbR, targeting the deletion area, was run on selected mutants to confirm clean iucABCD operon deletion.Similarly, a recombinant plasmid for the deletion of the part of the VirB4/D4 T4SS (virB1-virB8) was constructed by sewing upstream and downstream flanking regions of the virB1-virB8 genes by overlapping PCR using vir1/vir2 and vir3/vir4 primers.A SphI restriction site was included at the junction site of the two flanking regions to allow insertion of an antibiotic-resistance gene.The overlapping PCR product was cloned at BamHI and SalI sites of pDG1.Then, the hygromycin-resistant marker (hygR) amplified from pRGD-HmR (Addgene plasmid #74107, a gift from Hans-Martin Fischer) with hymf and hymr primers was inserted into the new SphI site on pDG1.This recombinant plasmid was transformed to strain SE163A.The allelic replacement procedure was performed as completed for the deletion of the iucABCD operon.Deletion of virB1-virB8 (replaced by hygR) was verified by colony PCR as above using vir5 and vir6 primers.

Construction of pDG3
The lambda t0 terminator was amplified from pDG1 using tse5 and tse7 primers.Salmonella's rhaBAD promoter including its rhaB ribosome binding site was amplified from strain SE163A using tse6 and tse9 primers.tse2, a type VII secretion system toxin encoding gene, was amplified from Pseudomonas aeruginosa PAO1 using tse8 and tse11 primers.The lambda tL3 terminator was amplified from pOSIP-KC using tse10 and tse12 primers.These four fragments were assembled at the NotI site of the pDG1 vector using HiFiMix, and the resulting vector was designated as pDG3 (Fig. 2).

Allelic replacement with pDG3
A recombinant plasmid for the deletion of the iucABCD operon was constructed in a pDG3 vector as described above.The allelic replacement procedure was performed as described for pDG1 with some modifications.Transformants were selected on LB plates supplemented with 0.5% glucose, X-Gal, and the required antibiotics as above.A single blue colony recovered from the 42°C incubator was grown with shaking at 30°C in LB supplemented with 1% rhamnose to an OD 600 of 0.2.This culture was washed with rhamnose (2%) supplemented M9 salts and plated on X-Gal containing M9 agar as described in reference (37).The iucABCD operon deletion was verified by PCR using iuc11 and iuc12 primers as described above.

Construction of pDG-Int
A DNA fragment containing ΦC31attP was amplified with ph1 and ph2 primers from pOSIP-KC (Addgene plasmid #138249, a gift from Drew Endy and Keith Shearwin) and cloned to the NdeI site on pKOVcir using HiFiMix.Then, the flippase recognition target (FRT), amplified from pKD3 (a kind gift of Matthew Jorgenson) using ph3 and ph4, was cloned with HiFiMix to the NotI site of the ΦC31attB containing vector to create pKOVcir-ΦC31attP-FRT.A region encompassing cmR, zeoR, ph31attP, and FRT was then amplified from this vector using gm1 and gm4 primers.Similarly, a region containing the R6K ϒ origin of replication was amplified from the pKD3 vector with gm2 and gm4 primers.The two fragments were then assembled by HiFiMix to create a new vector with the R6K ϒ origin of replication.Then, an integrase-encoding gene (int) amplified from pOSIP-KC using ph5 and ph6 was cloned to the NotI site of the vector with the R6K ϒ origin of replication using T4 DNA ligase.The resulting integrative vector was designated as pDG-Int (Fig. 3).

Construction of pDG-IntB
The bgaB gene, amplified using lac5 and lac6 (containing the trc promoter sequence) from pMAD vector, was TOPO cloned as above.Then, the TOPO-cloned bgaB gene was excised with EcoRI and cloned by T4 DNA ligase to the corresponding site on pDG-Int's cmR gene to create pDG-IntB (Fig. S3).

Construction of ΦC31attB/FRT knock-in recombinant plasmid
The pKD3 plasmid was linearized by fa1 and fa2 primers using inverse PCR.Each half of the ΦC31attB sequence was incorporated at the 5′end of these primers.The PCR obtained by fa1 and fa2 primers was circularized by T4 DNA ligase to generate a new vector containing the complete ΦC31attB site.Then, the ΦC31attB fragment, with its pre-existing FRT sequences from pKD3, was amplified by fa3 and fa7 primers.To knock in ΦC31attB for virD4, the upstream and downstream regions flanking the virD4 gene were amplified using fa4/fa5 and fa6/fa8 primers using SE163A as a template.These three fragments were then cloned at SalI/SacI sites on pDG1 with HiFiMix.To knock in ΦC31attB/FRT to the chromosomal rhaBAD promoter, upstream and downstream region flanking rhaBAD promoters were amplified by rha1/rha3 and rha4/rha6 primers using SE163A (Table 1) stain as a template.Then, a region containing ΦC31attB and FRT was amplified from ΦC31attB/FRT knock-in recombinant plasmid using rha2 and rha5 primers followed by the assembly of these three fragments to the pDG1 vector as above.To knock in ΦC31attB/FRT to the iucA promoter located on IncFIB plasmid, a region flanking this promoter was amplified by fib1/fib2 and fib5/fib6 primers using SE163A as a template.Then, the ΦC31attB/FRT fragment was amplified from DGS49 using fib3/fib4 primers.These three fragments were assembled to the SacI site on pDG3 vector as above.

Construction of ccdA/ccdB/FRT knock-in recombinant plasmid
ccdA/ccdB genes were amplified from strain SE163A using fa10 and fa13 primers.The FRT sequence was amplified from pKD3 using fa12 and fa15 primers.The virD2 gene upstream and downstream flanking regions were amplified using fa9/fa11 and fa14/ fa16 primers using SE163A strain as a template followed by the assembly of the four fragments to the pDG1 vector as above.

Construction of pDG-Int2 and pDG-Int3 vectors
The intand bgaB-encoding genes were inserted to pKOVcir-ΦC31attP-FRT to create pDG-IntB1 (Fig. S4) as described above.Then, the FLP-encoding gene along with its temperature-sensitive lambda repressor was amplified from pCP20 using flp1 and flp2 followed by cloning with HiFiMix at the ScaI site located at the 3′end of the bgaB-disrup ted cmR gene.The resulting vector was designated as pDG-Int2 (Fig. 4).To construct pDG-Int3, the rhaS gene was amplified from pSCrhaB2 (Addgene plasmid #113634: a gift from Miguel Valvano) using flp3 and flp5 primers.The lambda t0 terminator was amplified from pDG1 using flp4 and flp6 primers followed by cloning of these two fragments with HiFiMix at the ScaI site of pDG-IntB1 (Fig. S4).Then, the rhaB promoter was amplified from pSCrhaB2 using flp7 and flp9.Next, the FLP-encoding gene was amplified from the pCP20 vector using flp8 and flp10 primers followed by cloning of these two fragments with HiFiMix at the ScaI site of the rhaS carrying vector to generate pDG-Int3 (Fig. S9).

Integration of pDG-Int vectors
pDG-Int vectors were electroporated using the Salmonella strain containing the ΦC31attB site and selected on a media supplemented with zeocin and X-Gal as needed.pDG-Int3 transformants were selected on LB agar supplemented with 1% glucose, zeocin, and X-Gal.Integration into the chromosomal rhaBAD promoter was verified using rha7/zeo3.Excision of the pDG-Int backbone from the chromosome was verified by PCR amplification using rha2 and zeo4 primers.

Sucrose counterselection was not efficient in deleting genes encoded by Salmonella plasmids
pKOV was used as a backbone for construction of allelic replacement vectors.The pKOV vector confers resistance to chloramphenicol permitting allelic replacements in only chloramphenicol-susceptible strains (27).As many Salmonella carry multidrug resistance plasmids, including those conferring resistance to chloramphenicol, pKOV is not an ideal choice to knock out genes from multidrug resistance plasmids.To circumvent this challenge, we inserted a zeocin resistance marker into pKOV to form pKOV-ZeO.The iutA gene encodes ferric aerobactin receptor protein, and this gene is part of the siderophore synthesizing aerobactin operon along with iucABCD (9).We used pKOV-ZeO for an in-frame deletion of iutA from an IncFIB plasmid harbored by Salmonella enterica strain SE163A (5) using sucrose as a counterselection generating DGS4[SE163A(IncFIB∆iutA)]. One of the limitations observed was that sucrose counterselection was not very efficient in this strain, as most of the colonies that grew on sucrose retained the mutagenesis vector (data not shown).

Blue/white screening strategy can be used for scarless gene deletion in Salmonella plasmids
Bacillus stearothermophilus carries an X-Gal-hydrolyzing thermostable β-galactosidaseencoding bgaB gene that can be used for blue/white screening (38).pDG1 expressing the BgaB enzyme was developed as described above.To test this vector for allelic replacement, we completed an in-frame scarless deletion of an ~6-kb (iucABCD) region of the siderophore-encoding operon from the SE163A IncFIB plasmids (Table 1).Following the allelic replacement procedure, we could see white colonies on X-Gal-supplemented media (Fig. S5A), representing plasmid loss.Our PCR result showed that pDG1 success fully deleted this operon from each copy of the plasmid (Fig. S5B), indicating this vector can be used for scarless deletion of genes.The PCR failed to amplify an ~7-kb DNA fragment from the wild type, which is likely due to the inefficiency of taq polymerase in amplifying large DNA fragments.To this end, Phusion High-Fidelity DNA Polymerases were employed to successfully amplify the ~7-kb fragment from the wild-type strain (Fig. S5C).We further validated pDG1 by replacing part of an IncX4 plasmid's VirB4/D4 T4SS operon (virB1 to virB8) with a hygromycin resistance (hygR) gene generating DGS13.pDG1 can only be used for non-lactose-fermenting strains, or it requires deletion of lac operon from the host to allow blue/white screening in lactose-fermenting strains such as Escherichia coli and Klebsiella pneumoniae.Furthermore, pDG1 cannot be used to knock-in X-Gal-hydrolyzing enzymes that can, for example, be used as a reporter gene.To circumvent these drawbacks, we replaced bgaB in pDG1 with mCherry under the trc promoter to generate the pDG2 vector (Fig. S2).We then transformed E. coli and strain SE163A (Table 1) followed by incubation for 30 to 36 hours.Our result showed that both strains transformed with pDG2 were able to form pink colonies (Fig. S6A and  B), suggesting that mCherry can be used for pink-white screening as a tool to assess mutagenesis vector loss.pDG2 was validated by deleting the N-terminal sequence of an avian pathogenic E. coli's lacZ gene (39), and this deletion mutant (DGE209) appeared white on X-Gal-supplemented media as removal of the N-terminal sequence of lacZ can abolish its activity (40) (Fig. S6C and D).

Expression of an inducible toxin from pDG vectors can serve as a potent counterselection marker
Because pDG vectors do not express potent counterselection markers, recovery of white colonies, those that lost the allelic replacement vector, sometimes requires repeated attempt.As described by others (34)(35)(36)(37), vectors with appropriate counterselection can improve the success rates of mutant isolation.Therefore, we cloned tse2, a gene encoding type VII secretion system toxin (41), under Salmonella's rhamnose-inducible promoter (sP rhaB ) in pDG1 to generate pDG3 (Fig. 2).The native rhaB ribosome binding site was used for Tse2 expression.The choice of the Salmonella's rhaB ribosome binding site and promoter obtained optimum expression/repression of the toxin through the native rhaBAD expression system.Unintended growth due to sP rhaB 's "leaky" expression, if any, can be repressed through a process known as catabolic repression by growing the strain in a media supplemented with glucose (42).To determine if the Tse2 counter selection marker facilitated loss of the mutagenesis plasmid, we used pDG3 to delete the iucABCD operon from the IncFIB plasmid carried by strain SE163A.Subsequently, hundreds of white colonies on M9 agar supplemented with 0.2% rhamnose and X-Gal were visible (Fig. S7A).Well-separated white colonies were obtained in 10 −2 and 10 −3 dilutions that would otherwise be impossible with pDG1 (Fig. S7A).Screening of white colonies by PCR using primers flanking the iucABCD operon confirmed the deletion of the operon.As there were no big blue colonies detected on X-Gal-and rhamnose-con taining M9 agar (Fig. S7A), we reasoned that iucABCD operon deletion by pDG3 could be detected by colony PCR.To this end, we picked 14 random colonies and performed PCR as above.The results showed about 64% of the strains contained the desired mutants (Fig. S7B), indicating that gene deletion by pDG3 could be done without the need of X-Gal supplemented media.

pDG-Int failed to integrate to Salmonella plasmid with the ΦC31attB site
We developed an integrative vector with the R6K ϒ origin of replication, referred to as pDG-Int (Fig. 3), conferring resistance to chloramphenicol and zeocin.pDG-Int carries the ΦC31attP site, FRT-minimal, and ΦC31 integrase.This vector was expected to integrate the strains with the ΦC31attB site lacking the π-factor required for replication of plasmids with the R6K ϒ origin of replication.To evaluate if pDG-Int can integrate to Salmonella plasmids, we used the IncX4 plasmid where we replaced part of the VirB4/D4 T4SS with hygR as a target site (see above).To increase the stability of this plasmid, we knocked the CcdA/CcdB type II toxin-antitoxin system into the virD2 gene located upstream of the inserted hygR.Since pDG-Int can only integrate to a strain with a ΦC31attB site, we knocked in ΦC31attB to the virD4 gene located downstream from the inserted hygromycin hygR gene using pDG1.To facilitate FLP-mediated removal of the whole VirB4/D4 type IV secretory system along with the hygR gene, FRT was inserted into the virD2 and virD4 genes using pDG1.The existing FLP-expressing pCP20 plasmid confers resistance to ampicillin, and this vector may not be used in multidrug-resistant strains.Thus, we inserted the nourseothricin-resistant marker into pCP20 to generate pCP20-nrsR as described.As expected, introduction of pCP20-nrsR resulted in the removal of the whole VirB4/D4 system leaving behind FRT scar, attB site, and ccdA/ccdB genes generat ing DGS64 (Table 1).The removal of the VirB4/D4 system was verified using PCR and by virtue of their susceptibility to hygromycin (data not shown).We expected that pDG-Int can integrate into the engineered IncX4 plasmid via the attB site generating attR and attL sites that can serve as substrates for subsequent excision catalyzed by integrases and excisionase (43).However, repeated attempts of integrating pDG-Int into IncX4 failed.

pDG-Int can integrate to the ΦC31attB site on Salmonella chromosome without an off-site target
Plasmids, such as those in SE163A, may carry a R6K ϒ origin of replication or π-fac tor, thus impeding integration of the vector with the same origin of replication.In that case, pDG-Int could integrate into the chromosome of S. enterica strain SE819 which lacks plasmids (44).To elucidate this possibility, we first replaced the chromoso mal rhamnose operon promoter (P rhaBAD ) of the strain SE819 with FRT-attB to create SE819∆P rhaBAD ::FRT-attB (DGS49; Table 1).To take the advantage of blue/white screening in detecting FLP-mediated excision of pDG-Int after integration, we inserted bgaB by disrupting the cmR gene located on the pDG-Int to generate pDG-IntB (Fig. S3) so that FLP-mediated excision of the pDG-IntB would result in the removal of the bgaB gene, forming white colonies on media supplemented with X-Gal.To this end, transformation of pDG-IntB into DGS49 resulted in integration of the vector to the attB site without any off-site targeting as confirmed by PCR (data not shown).Interestingly, about half of the colonies formed white colonies on X-Gal supplemented media following transforma tion with pCP20-nrsR, indicating vector excision.Excision of the pDGint-bgaB backbone was further verified in the white colonies by PCR.pDG-IntB was able to integrate into the SE163A chromosomal rhamnose operon promoter, as in DGS49, and to the IncFIB plasmid carried by this strain, indicating that the lack of integration to SE163A's IncX4 plasmid was not due to the presence of a plasmid with the R6K ϒ origin of replication or its π-factor.Integration of this vector to the IncFIB plasmid and generation of attL and attR at each end was verified using int1/int2 and zeo4/int3 primers, respectively (Fig. S10).

pDG-IntB can be used to integrate as large as a 9.3-kb fragment to target the attB site
To evaluate the usefulness of pDG-IntB in inserting a larger fragment, we cloned iucABCD-iutA including its auxiliary shiF gene located upstream of this operon (45), into pDG-IntB followed by electroporation into DGS49.The following day, we observed few blue colonies and the subsequent introduction of pCP20-nrsR to the blue colonies resulted in the formation of white and blue colonies.Integration of iucABCD-iutA to DGS49's rhamnose operon and removal of vector's backbone were verified in white colonies using primers described above.
pDG-Int2 can integrate into the Salmonella chromosome and self-excise its backbone without the need of pCP20-nrsR.
To develop an integrative vector independent of the R6K ϒ origin of replication, we designed a new integrative vector (pDG-Int2) carrying a temperature-sensitive origin of replication (Fig. 4).In addition to ΦC31's attB and bgaB, pDG-Int2 carries an FLP-encoding gene with a temperature-sensitive lambda repressor as in the int gene.Repression of int and FLP by a lambda repressor can be relieved by incubating the strain at nonpermissive temperatures.To elucidate if pDG-Int2 can integrate into strain DGS49, we transformed pDG-Int2 and pDG-IntB1 (used as a control) into this strain followed by growth at non-permissive temperature.The next day, we observed many blue colonies for the control and few blue colonies for pDG-Int2 on X-Gal-supplemented media (data not shown).We suspect the blue colonies of pDG-Int2 transformants may contain a population of bacteria that have undergone plasmid backbone excision (Fig. 5).The PCR results confirmed that the blue colonies had undergone plasmid back bone excision.Interestingly, re-streaking the blue colonies on X-Gal-supplemented media formed a mixture of white and blue colonies, where the white colonies outnumbered the blue ones (Fig. S8).No white colonies were observed on the control plates.We checked for the complete removal of the plasmid backbone from the white colonies using PCR targeting the flanking regions outside of the integration site and found that the pDG-Int2 backbone was completely excised by FLP leaving behind zeoR and multiple cloning sites stably integrated into the chromosome (Fig. 5), but the number of blue colonies obtained following transformation was somewhat low.

Regulation of FLP expression in pDG-Int2 through the rhamnose-inducible promoter did not improve the number of blue colony transformants
Transforming the target strains with pDG-Int2 often generated few colonies or none.We wanted to determine whether tight regulation of FLP expression through the rhamnoseinducible promoter improves blue colony recovery.The FLP gene was cloned under P rha to generate the new vector pDG-Int3 (Fig. S9).We then transformed pDG-Int2 and pDG-Int3 to strain DGS49 (Fig. S9) and plated them on media supplemented with glucose and X-Gal.The following day, we observed few blue colonies in both pDG-Int2 and pDG-Int3 transformants indicating that repression of FLP via the rhamnose-induci ble promoter did not improve the efficiency of blue colony formation but transforma tion with pDG-Int3 gives a reproducible result compared with pDG-Int2.As expected, re-streaking those blue colonies on X-Gal supplemented media formed mixture of blue and white colonies and the excision of the vector was further verified by PCR as described above.

Cloning of the FLP gene to the pDG-IntB vector improved the number of blue-colony transformants
To evaluate whether the poor integration/excision efficiency observed in pDG-Int2/Int3 was due to its origin replicon, we cloned the FLP gene to pDG-IntB as described in pDG-Int2 construction.Electroporation of this vector to DGS49 generated many blue and white colonies (Fig. S11B); and re-streaking of the blue colonies formed blue and white colonies as observed in pDG-Int2.Integration and subsequent excision were confirmed by PCR as noted above.We next cloned the 9.3-kb fragment containing the iucABCD-iutA operon into pDG-Int4 as above followed by electroporation of the recombinant plasmid to DGS49.The next day, we could see few blue colonies, and as expected, re-streaking those blue colonies formed mixture of white and blue colonies with the white ones representing colonies with iucABCD-iutA integration and complete removal of vector's backbone (Fig. S11A).

DISCUSSION
Salmonella enterica is known to express a plethora of virulence factors which contribute to host immune invasion, intracellular survival, and pathogenesis (46,47).This species is also known to harbor multidrug resistance plasmids, some of which also carry virulence genes, spanning different incompatibility groups (3,6,7).Despite the development of excellent chromosomal gene deletion tools for Gram-negative pathogens (16)(17)(18)(19)(20)(21)(22)(23), efficient vector tools for the deletion of genes on plasmids, including those harbored by Salmonella and subsequent complementation of phenotypes related to the effect of gene mutation, are lacking.This limitation is because most of the suicide vectors are not efficient for plasmid genome mutations as they rely either on plasmid-derived counter selection toxins such as ccdB and relE (36,37,48) that can be neutralized by cognate antitoxin or other less efficient strategies such the use of tetA, rpsL or their combina tion, which often requires specific strain optimization (34).Additionally, in most cases, previously described suicide vectors contain the R6K ϒ origin of replication permitting gene deletion only in strains lacking plasmids with the R6K ϒ ori or the associated π-factor (31,37,49).Integrative vectors have not been described for gene complementation in Salmonella.Most integrative vectors are described in E. coli and used mainly for biotechnological applications (50,51), and even those lacking chromophore indicators require laborious screening or excision of a vector's backbone, which relies on the introduction of another FLP-expressing plasmid (15,52).This study documents vector tools that circumvent these drawbacks.The allelic replacement vectors described in this study contain a temperature-sensi tive origin of replications (27,32) that can be used to delete genes regardless of the plasmid replicon type that the strain is carrying.Sucrose counterselection during allelic replacement using pKOV-ZeO was not efficient likely due to sacB gene mutation as suggested by others (34)(35)(36)(37).Although not described for Gram-negative organisms, the bgaB gene has been used in vectors to generate markless gene deletion in low-GC content Gram-positive bacteria by allowing blue/white screening as an indicator for plasmid loss (53).Similarly, bgaB in our allelic replacement vector improved the success of mutant isolation by narrowing the screening procedure to those who lost the mutagenesis vector without a bias for those carrying it.In another similar study, blue/white screening via E. coli's lacZYA operon was used for allelic replacement in Salmonella, but this strategy is less practical as the size of the operon itself is over 5 kb compared with ~2 kb for bgaB; furthermore, colonies expressing lacZYA may not form a blue color at higher temperature in contrast to the thermostable BgaB enzyme (38).In recently described allelic replacement vectors, Lazarus et.al. (37) showed that singleand double-crossover events can be monitored using genes expressing blue AmilCP or magenta TsPurple non-fluorescent chromoproteins, but colony color saturation depends on the vector copy number or requires colony re-streaking (37), whereas the blue color formation via BgaB can easily be detected in individual transformants (Fig. S5).Alterna tively, gene deletions can be performed using pDG2 through a pink/white screening strategy both in lactose and non-lactose fermenters.It may be prudent to assess whether the strain forms pink colonies before performing the allelic replacement procedure.The Tse2 toxin has been used as potent counterselection in a two-step lambda recombinasemediated markless deletion in Enterobacteriaceae including in Salmonella (36) and in a newly described allelic exchange suicide vector (37).Similarly, the use of Tse2 as a counterselection in our allelic replacement vectors effectively suppressed cells carrying the mutagenesis vector.The use of the native rhaB ribosome for tse2 expression as suggested by Khetrapal et al. (36), and the choice Salmonella's rhaBAD promoter (Fig. 2), could have contributed to the high level of growth suppression.Any colony that escaped the deleterious effect of the toxin can easily be identified as they form a blue color on X-Gal-supplemented media.
A change in phenotype following gene mutation is usually complemented by expressing a copy of the target gene from plasmids, but this method suffers from several drawbacks including the use of antibiotics for plasmid maintenance, increased metabolic burden due to plasmid copy number, and multiple copy target gene expression that can potentially affect the outcome of the result.In this regard, integrative vectors are invaluable tools for generation of a highly stable single-copy complementing system.Gene integration using phiC3 integrase is one of the extensively used set of techniques in bacteria, plants, and mammalian cells (14,54,55) due to its stability and unidirec tional recombination (11,43,56).The integrative vectors described in this study, which expresses phiC3 integrase, have several advantages.Firstly, the vectors express the X-Gal hydrolyzing enzyme for tracking integration and excision via the blue/white screening, which avoids the time-consuming screening procedure.Secondly, these vectors can be used to insert relatively large fragments as evidenced by integration of a 9.3-kb fragment encompassing the aerobactin-synthesizing operon into Salmonella chromo some.Thirdly, excision of pDG-Int2 and pDG-Int4 backbone after integration is catalyzed by the FLP expressed from the integrated vector itself (Fig. 2 and 5), in contrast to other vectors where such excision including antibiotic resistance marker elimination requires introduction of another FLP-expressing plasmid (15,52).Self-excising vectors are particularly important for multidrug resistance strains where the use of a previously described FLP-expressing plasmid is impeded by limited antibiotic resistance marker use.In addition to these advantages, integration with pDG-Int2 and subsequent excision does not depend on the plasmid replicon type as compared with integrative vectors with the R6K ϒ origin of replication, where efficient integration requires a strain that lacks a plasmid with R6K ϒ ori or π-factor (15,50,51).
Nevertheless, coupling lambda red recombinase with the CRISPR/Cas9 or I-SceI system has been shown to be effective for chromosomal integration of fragments as large as 12 kb (26).Although this is an excellent tool for chromosomal insertion, we speculate that this system may not be efficient for inserting genes on a plasmid because of the lack of counterselection as double-strand breaks in plasmids are not lethal in contrast to chromosomal double-strand breaks that provide efficient counterse lection (19,21).Moreover, the aforementioned methods require considerable antibiotic resistance marker modification to create mutations in stains harboring multidrug-resist ant plasmids.These integrative vectors can also have potential applications beyond the use in clinically important bacteria, such as for heterologous expression of industrially relevant microbial natural products which are often found as multigene biosynthetic gene clusters where conventional knock in vectors are not efficient to insert such clusters (57).The low integration/excision efficiency observed for pDG-Int2/pDG-Int3 may emanate from its low copy pSC101 origin replication, as opposed to a vector with R6K ϒ ori, which is multicopy.It should also be noted that phiC31attB and FRT sites have been inserted to the target genome prior to integration procedure.
In summary, we developed versatile vectors that can be used to delete genes from plasmids and chromosomes as well as inserting relatively large fragments more efficiently than the conventional knock-in vectors.Furthermore, the vectors express chromophores for the detection of target gene modification or colony isolation, avoiding time-consuming screening procedures.These vectors can serve as an invaluable tool for the assessment of the role of plasmid/chromosomally encoded virulence factors for Salmonella pathogenesis.

TABLE 1
Strains and vectors used in this study