A large self-transmissible resistance plasmid from Nigeria contains genes that ameliorate a carrying cost

Antimicrobial resistance is rapidly expanding, in a large part due to mobile genetic elements. We screened 94 fecal fluoroquinolone-resistant Escherichia coli isolates from Nigeria for six plasmid-mediated quinolone resistance (PMQR) genes. Sixteen isolates harbored at least one of the PMQR genes and four were positive for aac-6-Ib-cr. In one strain, aac-6-Ib-cr was mapped to a 125 Kb self-transmissible IncFII plasmid, pMB2, which also bears blaCTX-M-15, seven other functional resistance genes and multiple resistance pseudogenes. Laboratory strains carrying pMB2 grew faster than isogenic strains lacking the plasmid in both rich and minimal media. We excised a 32 Kb fragment containing transporter genes and several open-reading frames of unknown function. The resulting 93 Kb mini-plasmid conferred slower growth rates and lower fitness than wildtype pMB2. Trans-complementing the deletion with the cloned sitABCD genes confirmed that they accounted for the growth advantage conferred by pMB2 in iron-depleted media. pMB2 is a large plasmid with a flexible resistance region that contains loci that can account for evolutionary success in the absence of antimicrobials. Ancillary functions conferred by resistance plasmids can mediate their retention and transmissibility, worsening the trajectory for antimicrobial resistance and potentially circumventing efforts to contain resistance through restricted use.


Results
Plasmid-encoded quinolone resistance genes in fluoroquinolone resistant isolates from nigeria. Of 1,098 fecal E. coli isolates originally recovered from 134 mother-infant pairs, 94 (8.6%) were ciprofloxacin-resistant and were screened for six PMQR genes. The PCR screen initially revealed that aac(6′)-Ib was common among ciprofloxacin resistant isolates with 37 (39.3%) being positive for the gene as shown in Table 1. However, upon BstI restriction analysis necessary to confirm the ciprofloxacin resistance-conferring allele, only four of the isolates with aac(6′)-Ib genes (10.8%) proved to be the aac(6′)-Ib-cr allele (Table 1). qnrA and qnrB were not detected at all whilst qnrS1, which we have previously detected in a different strain set from Nigeria 19 , was found in two of the isolates. Overall, oqxAB was detected in nine isolates (including two that were positive for aac(6′)-Ib-cr) and was therefore the most common PMQR recorded but aac(6′)-Ib-cr was the most common quinolone-specific mechanism (Table 1). We therefore went on to characterize M63c, one of the strains bearing aac(6′)-Ib-cr alleles.
Strain M63c was mucoid on MacConkey and eosin methylene blue plates with an API 20E profile pointing to 'group C E. coli' , which are now classified as Escherichia fergusonnii. We amplified and sequenced the 16S rRNA gene of M63c and found the sequence (Accession number SUB5013267) was 99% identical to that of Escherichia fergusonii type strain ATCC 35469 (Accession number NR_074902.1). However, Maheux et al. 20 reported that the adk, gyrB and recA genes were more useful for speciating Escherichia spp. than 16S rRNA. Alleles of those genes carried by M63c were identical to E. coli alleles. We additionally sequenced fumC, icd, mdh and purA genes of the strain and multilocus sequence typed the isolate as an E. coli ST167 strain. E. coli ST167 strains are encountered around the globe, frequently multiply resistant and often expressing extended-spectrum β-lactamases (ESBLs) 21 . We extracted a large plasmid DNA from strain M63c and electroporated it into DH5α cells (DH5αE from Thermo Scientific). Transformants were obtained on neomycin (50 mg/ml) plates but not on ciprofloxacin (1 mg/ml). *aac(6′)-Ib 26 37 aac(6′)-Ib-cr 2 4 Table 1. PMQR genes identified by PCR in ciprofloxacin non-susceptible Escherichia isolates. *Includes the PMQR aac(6′)-Ib-cr as well as non-ciprofloxacin resistance conferring aac(6′)-Ib.
Each transformant showed reduced susceptibility (but not full resistance) to ciprofloxacin and nalidixic acid as well as resistance to tetracycline, ampicillin, neomycin/kanamycin, trimethoprim and sulphonamides but not chloramphenicol. All transformants had identical plasmid DNA BamHI, and SalI restriction profiles and susceptibility patterns and carried a single large plasmid, which we termed pMB2.
Sequence of Plasmid pMB2. pMB2 was Sanger-sequenced and shot-gun assembled into a 125,782 base pair circular replicon (Genbank Accession number MK370889). pMB2 has a G + C content of 51.73% (Fig. 1), which is comparable to that of the Escherichia coli chromosome 22 A maximal number of 323 ORFs of approximately 100 or more base pairs in length were predicted but additional evidence supports only 145 of these as encoding real or hypothetical genes. The plasmid contains a full IncFII-type conjugation region of 32 genes including all 25 of the consecutive transfer region genes with known functions in the conjugation process 23 .
In silico replicon typing revealed that the Plasmid Sequence Type of pMB2 is [F36:A4:B1]. The pMB2 plasmid carries replication-associated proteins associated with IncFIA, IncFIB, IncFIC and IncFII plasmids. The IncFIC RepA1 protein is coded by an ORF at 86,659 bp on the plasmid, which is close to the major shift in G + C skew seen at 81,000 bp and likely marking the origin of replication (Fig. 1). pMB2 is rich in mobility genes with at least 26 transposases or transposase-like ORFs, as well as up to six other integrases and recombinases encoded. Almost all functional units on the plasmid are flanked by insertion sequences. pMB2 carries genes encoding several toxin-antitoxin systems ccdA/ccdB, vapB/vapC and pemI/pemK. It also carries genes encoding the SOS response inhibiting proteins PsiA and PsiB. Between positions 88,324 and 93,566 is a complete, six gene arginine deaminase operon, which could potentially catabolize arginine to ammonia.
pMB2 is predicted to encode at least three transport systems. The EamA-like domain-containing drug/metabolite and carboxylate/amino acid/amine transporter (Pfam00892; located at 107699-108583), is found in a variety of integral membrane proteins; there is no known function for these genes but they could possibly contribute to antimicrobial resistance. A hypothetical divalent ion transport system (located between 10,045 and 7,247, on the complementary strand) is composed of two proteins coded in an operon: conserved hypothetical pMB2_00012 a (pfam10670) 24 and HmuR/hemR, which is a Ton-B dependent hemoglobin-hemin receptor localized to the bacteria's outer membrane 25 . The third transport system, and the best studied, is encoded by sitABCD located on the forward strand of the plasmid between 19,195 and 22,644. SitABCD form an effective ferrous iron uptake system earlier described in Salmonella enterica and Shigella 26,27 . SitA is a periplasmic binding protein, SitB is the ATP-binding component, SitC functions as a permease and SitD is the inner membrane component of the system. pMB2 bears eight genes conferring resistance to six different drug classes (Fig. 2). An IS26-composite transposon flanks an incomplete class I integron carrying a dfrA17 gene (encoding resistance to trimethoprim) and an incomplete aadA5 gene. The tetracycline resistance gene tetA (alongside tetR, encoding the repressor of the tetracycline resistance element) is located within the remnants of a Tn1721 transposon. Next to this region, another ORFs in black and a BLASTn comparison between pMB2 and the five best hits E. coli plasmids retrieved from NCBI (from the innermost to the outermost circles, plasmids p6409, pECAZ146_1, p109, pRCS52 and pRCS57 with the GenBank accession numbers CP010372.1, CP018990.1, CP023372.1, LO017736.1 and LO017738.1, respectively). A brown star marks the likely location of the origin of replication.
IS26-composite transposon was identified, carrying an incomplete class I integron with a |aadA16|aac(6′)-Ib-cr|bla OXA-1/30 |ΔcatB3| gene cassette array encoding resistance to aminoglycosides, fluoroquinolones, beta-lactams and phenicol. As stated previously, the aac(6′)-Ib-cr gene, also named aacA4-crC, confers resistance to aminoglycosides and the broad-spectrum antimicrobial ciprofloxacin due to a T to C mutation at nt 283 and a G to T mutation at nt 514. Following bla OXA-1 is catB3, which normally inactivates chloramphenicol through acetylation 28 but which is not active in pMB2 due to a truncated cassette region. Next to the IS26-composite transposon is an aac(3)-II gene, conferring gentamicin resistance. Finally, a transposition unit carrying bla CTX-M-15 , which confers resistance to oxyiminocephalosporins, and ISEcp1 was found within the remnants of a Tn2 transposon. This insertion unit is flanked by 5-bp direct repeats (5′-TATGA-3′), suggesting the en bloc transposition of these elements within the Tn2 transposon (Fig. 2).
pMB2 is similar to many other resistance plasmids in the database that have been isolated from around the world. Figure 1 compares the sequences of four of these plasmids with pMB2 illustrating that the plasmids are almost completely conserved and almost perfectly syntenic. In particular, conjugation and arginine deaminase genes are highly conserved among the plasmids whereas their resistance regions are expectedly the most variable. Only three of the plasmids carry all the putative transport genes, however all of them have at least two of them. pMB2 is self-transmissible. While the plasmid as a whole is replete with transposable elements and insertion sequences, and contains multiple antimicrobial resistance genes, some no longer functional, others redundant, the conjugation region of pMB2 is not interrupted. The transfer genes of pMB2 are organized similarly to those of the F plasmid 29 and identical to those on the other five plasmids we compared with pMB2 in Fig. 1. Conjugation from DH5α (pMB2) to EC1502 (rifampicin resistant) on solid media occurred at the rate of 1.8 × 10 −5 to 8 × 10 −6 .
Three other Escherichia isolates from the individual from whom strain M63c was isolated had been archived along with two isolates from her infant 17 . As shown in Table 2, strain M63c was the only quinolone-non-susceptible isolate of this set. We screened all six isolates for the presence of the aac(6′)-Ib-cr allele and pMB2 and found that they were present in strain M63b, as well as original source strain M63c. While they differed in quinolone resistance, strain M63b and M63c shared otherwise identical susceptibility patterns. We sequenced quinolone resistance determining regions of gyrA and parC in both strains and found that quinolone-susceptible strain M63b contained no SNPs whilst resistant strain M63c carried Ser83Leu and Asp87Asn mutations in GyrA and Ser80Ile, Glu8Val mutations in ParC. These findings are consistent with reports that demonstrate that those SNPs are sufficient for high-level fluoroquinolone resistance and aac(6′)-Ib-cr allele is insufficient to confer clinically significant ciprofloxacin resistance but augments resistance conferred by other mechanisms 12,30 . The presence of the plasmid in two different Escherichia strains from the same host strongly suggests in vivo transmission. pMB2 does not have a carrying cost. We hypothesized that the large size of pMB2 would slow the growth of laboratory strain DH5α. In rebuttal of our hypothesis, DH5α carrying pMB2 grew at rates that were comparable to or faster than the plasmid-free strain. This occurred in Luria broth (LB) (Fig. 3) as well as in nutrient broth, www.nature.com/scientificreports www.nature.com/scientificreports/ terrific broth, Dubelcco's Modified Eagle Medium (DMEM, Invitrogen), Davis minimal medium and M9 minimal medium (data not shown). As a control, we used a 92 kb aggregative adherence impAB plasmid, pLMJ50 31 , which did demonstrate a significant carrying cost in DH5α (Fig. 3).

Characterization of pMB2.
To determine whether nutrient acquisition conferred by the pMB2 region harboring transport systems operons pMB2_00012/hemR and sitABCD located at 7,247-8,038 bp and 19,195-22,644 bp respectively on pMB2 offsets the carrying cost of pMB2, this segment of the plasmid was excised. A 32,331 bp segment that included these genes was deleted from the plasmid by excision with Not1 (NEB), which cuts pMB2 at 122,777 bp, and Xba1 (NEB), which cuts pMB2 at 29,449 bp. The restricted ends were filled in using E. coli DNA polymerase I (NEB) and then ligated on one other to generate a 93,450 bp autonomously-replicating  Table 2. Antimicrobial susceptibility profiles of Escherichia isolates from the mother-infant pair from whom plasmid pMB2 were recovered and DH5α with and without pMB2. Strains M63a-e are from a mother enrolled in the study of Odetoyin et al. 17 and strains C63a and c are isolates from her child. www.nature.com/scientificreports www.nature.com/scientificreports/ mini-plasmid, pRMKO. The excised 32,331 bp fragment containing the transport system region was cloned into the XbaI and NotI sites of pBluescript II SK (Agilent) and named pRMC.
As seen in Fig. 4, the LB growth data, the equivalent or faster growth conferred by pMB2 was not seen with the mini-plasmid pRMKO, which grew at a rate that was slower than equivalent-sized plasmid pLMJ50. Intriguingly, pRMC, the cloned deleted region also conferred slower growth than wildtype, although in this case the carrying cost may be attributed to the higher copy number of the clone. Attempts to move the large insert to a lower copy-number vector were not successful and therefore it was not possible to fully complement the mini-plasmid deletion in trans. We therefore elected to functionally characterize the sitABCD genes that are within the deleted sequence.
We amplified the sitABCD operon and its promoter from plasmid pSIT1 (which has an identical copy of the genes) received from Laura Runyen-Janecky 26 , using primers sitF and sitR and cloned them into pACYC177 to create pINK2301. In Davis minimal media iron-depleted by EDDA, DH5α grows much slower than M63c, the E. coli strain from which pMB2 was isolated. Transformation of DH5α with pMB2 produced a faster growth rate that was lower in DH5α (pRMKO). As shown in Fig. 5, this growth defect could be complemented in trans with pINK2301.
To determine whether pMB2 enhanced bacterial fitness, we conducted competition experiments between strain M63c, the strain from which pMB2 was originally isolated, and commensal E. coli strain HS. We also competed DH5α strains carrying pMB2 and pRMKO against one another. These experiments were performed in rich media (LB) as well as in DMEM in the absence of antibiotics, however antibiotics were used to select competitors on plates to perform counts. In both media, strains carrying pMB2 outcompeted commensal or laboratory strains. Conversely, DH5α (pRMKO) suffered a competitive disadvantage against DH5α (pMB2), as well as against plasmidless DH5α (Fig. 6).
We sought to determine whether the plasmids were stably inherited and found that this was the case in source strain M63c over 200 generations. However, pMB2 and pRMKO were less stably inherited in a DH5α background, with pRMKO proving to be less stably inherited than pMB2 in the medium term (Fig. 7).

Discussion
In this study, we identified and characterized pMB2 a multidrug resistance plasmid extracted from an Escherichia strain isolated in Nigeria. We were drawn to the plasmid by its aac(6′)-Ib-cr gene but its resistance region also includes genes encoding resistance to seven other classes of antimicrobials, including the ESBL gene bla CTX-M- 15 . There have been several reports of co-occurrence of bla CTX-M- 15 and aac(6′)-Ib-cr across of the globe, including in www.nature.com/scientificreports www.nature.com/scientificreports/ Nigeria 32-35 . This common co-resistance profile implies that ciprofloxacin use, which is growing 1 , also selects for the carriage of ESBL genes and vice versa explaining why IncFII plasmids carrying these genes have traditionally been associated with successful pandemic clones 36,37 . This report appears to be the first one directly confirming a genetic linkage of bla CTX-M- 15 and aac(6′)-Ib-cr in Nigeria. However the structure of this plasmid in general and the resistance region in particular has been reported from round the world.
When resistance is conferred by a plasmid of considerable size, the associated metabolic burden on the organism due to the additional energy required in replication and gene expression typically exerts a carrying cost, at least theoretically 38 . In practice, this is not always true but the co-evolutionary routes to reduced cost are often unknown. Some research has suggested that this burden is reduced or eliminated through compensatory mutations or acquisitions 14,[39][40][41][42][43] . Alternatively, or in addition, plasmid-encoded resistance genes themselves could confer a fitness advantage in the absence of antimicrobials 14,41 .
Antimicrobial use is the principal driving force behind the current epidemic of antimicrobial resistance. However, research has shown that, once evolved, multidrug resistance plasmids can be maintained in the absence of selection 44 . There is a dearth of fitness defects reported for resistance plasmids in the literature and the reasons why some of them are evolutionarily successful are not well understood 41,45 . This study demonstrates that the aac(6′)-Ib-cr gene of strain M63c is carried on a large multidrug resistance plasmid pMB2, which also bears a number of genetic loci that support its evolutionary success irrespective of antibiotic selection. The 125 Kb plasmid does not have a carrying cost in vitro and competition experiments verified that 32 Kb region of the plasmid accounts for selective success. Included in this evolutionarily significant 32 Kb region are the sitABCD genes, which promote growth under iron-limited conditions, along with as yet unidentified loci that promote growth in rich and minimal media. The sitABCD and adjacent genes studied in our research are highly conserved among most of them and therefore the fitness they confer is also likely to be common. Overall, these data suggest that fitness conferring mobile elements associated with resistance can co-evolve with host bacteria to compensate the costs of their carriage and propagate its transmission irrespective of antibiotic selection. While in vivo fitness is yet to be evaluated for these loci, and should be an area of future study, the in vitro advantages conferred were substantial. Without a carrying cost large multidrug resistance plasmids could persist in commensal E. coli even in the absence of antimicrobial pressure. Future work on resistance plasmids and elements should focus more intently on fitness genes in addition to those that encode drug resistance.   Table 3. Commensal E. coli were collected from Nigerian mother-infant pairs where the infant suffered from diarrhea as previously described 17 . Altogether stool samples from 134 mothers-infant pairs, yielded 1,098 E. coli isolates for evaluation. Routine culture of bacteria was performed aerobically at 37 °C in Luria broth supplemented with chloramphenicol (30 μg/mL), ampicillin (100 μg/mL), tetracycline (25 μg/mL), or neomycin (50 μg/mL) where applicable. Strains were maintained at −70 °C in Luria broth:glycerol 1:1.

Strains. Strains used in this study are listed in
Antimicrobial susceptibility testing. Antimicrobial susceptibility testing was performed using the Clinical and Laboratory Standards Institute (CLSI) disc diffusion method and interpreting zone diameters in accordance with CLSI guidelines in WHONET software version 5.3 46 . Discs used contained ampicillin (10 µg/ml), streptomycin (10 µg), trimethoprim (5 µg), tetracycline (30 µg), nalidixic acid (30 µg), chloramphenicol (300 µg), sulphonamide (1000 µg) and ciprofloxacin (5 µg) (Oxoid/Remel). E. coli ATCC 35218 and DH5αE (Invitrogen) were used as control strains. Minimum inhibitory concentrations (MICs) for nalidixic acid and ciprofloxacin were determined using E-test (bioMérieux) on Mueller-Hinton (MH) agar. Kanamycin MICs were determined by agar dilution on Mueller-Hinton Agar. Naturally-occurring low copy number larger plasmids were extracted, after growth of host bacterium in Terrific Broth and induction with chloramphenicol, by a modified boiling protocol for small scale preparations 47,48 and using the Qiagen ® Large Construct Kit for large-scale preparations. Table 4 lists the plasmids used in the study, their relevant properties, and their sizes. Plasmids were electroporated into E. coli host strains using a Bio-Rad micropulser according to manufacturer's instructions. Sequences of oligonucleotide primers used in the study are listed in Supplementary Table 1. Amplification cycles began with a two-minute hot start at 94 °C followed by 30 cycles of denaturing at 94 °C for 30 s, annealing at 5 °C below the melting temperature (unless otherwise specified) for 30 s, and extending at 72 °C for one minute per kilobase of DNA to be amplified. BstCI restriction analysis of aac(6′)-Ib PCR amplicons were used to determine whether they represented the ciprofloxacin resistance conferring aac(6′)-Ib-cr allele as it only differes by two SNPs and is therefore indistinguishable by PCR 49 . Up to two amplicons from each unique profile were sequenced to confirm PCR-RFLP-based classification. Where necessary, PCR amplicons were TA-cloned into the pGEMT vector (Promega) according to manufacturer's directions and plasmids were transformed into chemically competent E. coli K-12 TOP10 cells for sequencing. Other molecular biology operations were performed using standard procedures 50 .

Bacterial identification.
Bacteria were identified to the species level as described previously 17 and further biotyped on the API20E system (bioMérieux). Genetic identification was performed by 16S ribosomal subunit sequencing, amplifying the gene encoding M63c's 16S rRNA with primers 10 and 1507R (Supplemental Table 1) in a thermocycler using the cycle at 94 °C for 2 m, [94 °C for 30 s, 59 °C for 45 s, 72 °C for 90 s] x36, and 72 °C for 10 m 51 . Multilocus sequence typing (MLST) was performed by amplifying the adk, fumC, gyrB, icd, mdh, purA and recA genes using the primers of Wirth et al. as previously described 52,53 . PCR amplicons for sequencing were size-verified by agarose electrophoresis, cloned into pGEMT, and electroporated into DH5αE (Thermo Scientific) using a Biorad Micropulser. Plasmids were extracted with a Qiagen Plasmid Mini Kit and Sanger sequenced from M13F and M13R priming sites.
Shot-gun sequencing and sequence analysis. Whole-replicon shotgun library preparation, Sanger sequencing and assembly of plasmid pMB2 was performed by SeqWright DNA Technology Services (Houston, TX). Preliminary sequence analyses and annotation was performed manually by four authors in Artemis 54 with open reading frames initially predicted using Glimmer [55][56][57] and basing gene annotations on 98% or greater identity at the nucleotide and amino acid levels if BLAST e-value was 0 58 . Plasmid MLST typing was performed in silico, using PlasmidFinder and pMLST versions 2.0 59 . Open Reading Frame (ORF) identity was determined using BLAST and Pfam 56,57 . Two automated annotations were performed using BASys 55 and Prokka v 1.13.3 60 . Five authors resolved annotation discrepancies among automated and manual annotations. Antimicrobial resistance genes and their genetic context were searched using Galileo AMR (https://galileoamr.arcbio.com/mara/) 61 and ResFinder 58 . Direct and inverted repeats were identified by dot-plot analysis of pairwise FASTA alignments made using the BLAST suite. ORF and feature plots were prepared using BLAST ring image generator 62 . In vitro conjugation. In vitro conjugation experiments were performed by solid-surface matings. Donors and recipients were cultured in LB with appropriate selective antimicrobials. 0.5 mL of donor and recipient culture, grown overnight with selection, was spun down at 5000 rpm for 5 minutes in a fixed rotor microcentrifuge and resuspended in 20 μl of LB without antibiotics. The suspension was spotted onto dried LB plates, allowed to dry at room temperature for 15 minutes and then incubated at 37 °C for three hours. The mating reaction was resuspended in 1 ml of LB with vortexing and placed on ice to terminate conjugation. After mating, serial ten-fold dilutions of each terminated reaction was made in cold phosphate buffered saline and plated onto plates containing tetracycline (or other appropriate antimicrobials for controls)-to select for the plasmid-and nalidixic acid-resistance to which is conferred chromosomally in the recipient. Transconjugant colonies were counted after overnight incubation at 37 °C and up to 50 colonies were verified by plasmid profiling, phenotype on MacConkey and Eosin methylene blue agars, PCR-RFLP for the fliC allele 63 , and PCR for donor-and  Growth curves. Growth curves were plotted following growth of each strain in microtiter wells containing 160 µl of media in an Infinite ® 2000 Pro Series microplate reader preset at 37 °C (Tecan). Commercially available medias were prepared as recommended by the manufacturers. Iron-depleted media was prepared by adding deferrated ethylene diamino-o-dihydroxyphenyl acetic acid (EDDA) as described previously 65 . Test strains were first cultured overnight in 5 mL of LB media containing appropriate antibiotics where required. Overnight cultures were spun down, resuspended in 5 mL of 1X Phosphate Buffered Saline (PBS), and diluted to an optical density of 0.70 at 595 nm. 10 µl of the diluted cultures were then loaded into assigned wells in the microtiter reader. Absorbance was measured at 595 nm every 30 min preceded by 30 sec of orbital shaking for 24 hours. The average optical density of 25 reads per well at each time point was used in the analysis and up to six wells were tested per strain. Each specific strain and conditions was tested on a minimum of three occasions. plasmid stability. Plasmid stability was assessed by serially passaging E. coli strains carrying pMB2 and its derivatives as described by Sandegren et al. 66 . Triplicate starting cultures were grown overnight at 37 °C in 1 mL of LB supplemented with ampicillin or tetracycline. Bacterial cells were washed, resuspended in LB without antibiotics and an aliquot of 100 µL of washed cells was inoculated into 10 mL of LB in a 50 mL conical tube and incubated overnight at 37 °C. 100 µL of this overnight culture was serial passaged to 1 mL of LB daily, approximating 10 generations of growth per passage. At select time points, samples were diluted and plated onto plain MacConkey plates and MacConkey plates containing tetracycline.
In vitro competition experiments. We measured relative fitness in traditional competition assays as outlined by Wiser and Lenski 67 . Fitness was computed independently in LB broth and DMEM and E. coli strains carrying pMB2 and its derivative plasmids were selected using antibiotic markers present on those plasmids. Briefly, separate cultures of the competing strains were grown overnight in the test medium containing ampicillin at 37 °C before volumes equivalent to equal optical densities were added to a 50 mL conical tube to a final volume of ≈0.1 ml, which was immediately made up to 10 mL with test medium. The culture was mixed, 100 µL was withdrawn for viable counting (diluting with normal saline) and the culture was then incubated at 37 °C with shaking. At 24 h, 100 µL was withdrawn, diluted with normal saline, and plated out for counting. For wildtype strains, counting was performed on MacConkey agar plates containing ampicillin, tetracycline and no antibiotics. For strains carrying pMB2 or its mini-plasmid deletion construct, pRMKO, counting was performed on MacConkey agar plates containing ampicillin, Mueller Hinton tetracycline plates and Mueller Hinton trimethoprim plates. Differential computation of strains carrying pMB2 or pRMKO were made possible by the fact that pMB2's dfrA17 gene, conferring trimethoprim resistance, is deleted from pRMKO. Fitness (w) was calculated by comparing the ratio of competed bacterial strains obtained at 24 h, A f and B f hours respectively, with those computed from the sample drawn at0 h (A i and B i , respective), that is, using the formula specified by Wiser Table 4. Plasmids used in this study.