Article

# Multicopy plasmids potentiate the evolution of antibiotic resistance in bacteria

• Nature Ecology & Evolution 1, Article number: 0010 (2016)
• doi:10.1038/s41559-016-0010
Accepted:
Published online:

## Abstract

Plasmids are thought to play a key role in bacterial evolution by acting as vehicles for horizontal gene transfer, but the role of plasmids as catalysts of gene evolution remains unexplored. We challenged populations of Escherichia coli carrying the blaTEM-1 β-lactamase gene on either the chromosome or a multicopy plasmid (19 copies per cell) with increasing concentrations of ceftazidime. The plasmid accelerated resistance evolution by increasing the rate of appearance of novel TEM-1 mutations, thereby conferring resistance to ceftazidime, and then by amplifying the effect of TEM-1 mutations due to the increased gene dosage. Crucially, this dual effect was necessary and sufficient for the evolution of clinically relevant levels of resistance. Subsequent evolution occurred by mutations in a regulatory RNA that increased the plasmid copy number, resulting in marginal gains in ceftazidime resistance. These results uncover a role for multicopy plasmids as catalysts for the evolution of antibiotic resistance in bacteria.

Plasmids mediate the horizontal gene transfer (HGT) of accessory genes between bacteria, making them important drivers of prokaryotic evolution1,2. For example, the horizontal transfer of antibiotic resistance genes has played a very important role in the evolution of resistance in pathogenic bacteria. Although the benefits of HGT are obvious, it is still not understood how plasmids persist in bacterial populations. Classical theoretical models predict that HGT through conjugation is necessary for the maintenance of plasmids, given the fitness cost of plasmid carriage3,4 and the potential for spontaneous plasmid loss during cell division5,​6,​7. More recent analyses demonstrate that compensatory adaptation alleviates the fitness cost due to plasmids8,​9,​10, allowing plasmids to stably persist for hundreds of generations, even in the absence of HGT9,11. However, the regimes of positive selection for plasmid-encoded genes that are necessary for the long-term stability of plasmids are stringent11. Finally, it is possible for plasmid genes to be transferred to the bacterial chromosome, for example by transposons, allowing the bacterium to lose costly plasmids while maintaining beneficial genes12. Therefore, the ‘plasmid paradox’ of how plasmids persist in bacterial populations in the long term is not completely solved, especially for small multicopy plasmids that lack maintenance systems such as active partitioning or post-segregational killing systems. However, these small plasmids are extremely prevalent in nature and in hospitals, as has been recently shown in E. coli, where almost 90% of ST131 strains carry small ColE1 plasmids13.

One possible solution for this paradox could be that certain genes confer a larger benefit to the bacteria when encoded on plasmids compared with on the chromosome. For example, carrying a gene on a plasmid with multiple copies produces higher levels of expression, and this may be beneficial for the host14. Another possible advantage is the variation in plasmid copy numbers that exists among single cells, which generates heterogeneity in gene expression between plasmid-bearing bacteria, leading to potentially beneficial phenotypic plasticity15. Finally, it is also possible that multicopy plasmids could accelerate the evolution of the genes they carry by increasing gene mutation rates due to the increase in gene copies per cell16,​17,​18,​19. Under this hypothesis, multicopy plasmids could work as platforms for gene variation and evolution, facilitating the adaptation of the host bacteria to new selective pressures.

Here we investigate the evolutionary advantages associated with carrying a beneficial gene, in this case an antibiotic resistance gene, on a multicopy plasmid versus in the chromosome.

## Results

### Experimental system

We constructed an experimental model using E. coli strain MG1655 (MG) and the β-lactamase gene blaTEM-1. Hundreds of variants of TEM-1 exist that provide resistance to different β-lactam antibiotics (http://www.lahey.org/studies) and they are usually encoded on plasmids, including small multicopy plasmids20,​21,​22. In addition, TEM-1 is one of the best-characterized systems for the study of molecular evolution in the laboratory17,23,​24,​25. We used TEM-1 as a model system to explore the potential of multi­copy plasmids as catalysts of gene evolution in general, and of the evolution of antibiotic resistance in particular. We inserted blaTEM-1 in the bacterial chromosome to generate E. coli MG1655::blaTEM-1 (MG::blaTEM-1 hereafter, see Methods). In parallel, we constructed a non-transmissible multicopy plasmid carrying blaTEM-1 with an identical promoter (Fig. 1a). This plasmid also carries a gfp gene under the control of an inducible l-arabinose promoter, the gene coding for the repressor of this promoter araC, and a chloramphenicol resistance acetyltransferase gene, cat. We named this plasmid pBGT and transformed it into E. coli MG1655 generating the strain MG/pBGT. This plasmid had 19 copies per bacterium (average = 19.1, s.d. = 3.8, n = 6), which is similar to the copy number of other small natural multicopy plasmids26,27. This experimental system allowed us to evaluate the effects of a plasmid versus chromosomal location of blaTEM-1 in isogenic strains.

### Plasmid and chromosome blaTEM-1-mediated resistance

Previous studies have shown that increasing the dosage of the blaTEM-1 gene produces an increase in β-lactam resistance28,​29,​30, suggesting that carrying blaTEM-1 on a plasmid is likely to increase antibiotic resistance. TEM-1 confers resistance to ampicillin, so we measured the minimal inhibitory concentration (MIC) of ampicillin in the different strains (Table 1). The MIC for the susceptible parental MG strain was 4 mg l−1 and it increased to 512 mg l−1 in MG::blaTEM-1 and to 8,192 mg l−1 in MG/pBGT. Therefore, even if blaTEM-1 conferred high-level ampicillin resistance both in the chromosomal and plasmid locations, the multicopy plasmid-bearing strain has a 16-fold increase in MIC compared with the one carrying blaTEM-1 in the chromosome.

Although resistance plasmids provide a fitness benefit in the presence of antibiotics, they typically impose a fitness cost in the absence of antibiotics3,4. We used competition experiments to measure the relative fitness of the blaTEM-1-carrying strains (Fig. 1b). The presence of blaTEM-1 in the chromosome produced no cost in MG in the absence of antibiotics (relative fitness (W)  = 0.997, two-sample t-test, P = 0.343, t = 1, d.f. = 9), suggesting that the effect of blaTEM-1 insertion in the chromosome was negligible. To study the fitness effects of pBGT, we competed the strains carrying blaTEM-1 on the chromosome (MG::blaTEM-1) or on the plasmid (MG/pBGT) under a range of ampicillin concentrations (Fig. 1b). In the absence of antibiotics, pBGT was associated with a 5.6% reduction in fitness (one-sample t-test, P < 0.001, t = −9.35, d.f. = 13), which is similar to the cost produced by natural multicopy plasmids observed in previous studies4,26. However, at concentrations of ampicillin above 128 mg l−1 the benefits of carrying pBGT exceeded its costs, and pBGT carriage became highly beneficial at 256 mg l−1. This result highlights the dynamic fitness effects of plasmids depending on the environmental conditions, producing associations with the host bacteria that range from parasitic to symbiotic10,31.

### Plasmid location of blaTEM-1 potentiates the evolution of ceftazidime resistance

TEM-1 produces high-level resistance to ampicillin but it has minimal activity against later β-lactam antibiotics. For example, the MIC of ceftazidime, a third-generation cephalosporin, was similar in the parental MG strain (0.25 mg l−1) and in strains carrying TEM-1 (MG::blaTEM-1 0.25 mg l−1; MG/pBGT 0.5–1 mg l−1). One possible benefit conferred by multicopy plasmids could be an increased evolvability of the genes they encode due to the higher number of gene copies per cell. In this case, it is well known that mutations in blaTEM-1 can expand the range of activity of TEM-1 to hydrolyse cephalosporins much more efficiently32, suggesting that pBGT could potentiate the evolution of novel variants of TEM-1. To test this hypothesis, we compared the ability of populations of MG::blaTEM-1 and MG/pBGT to evolve resistance to high doses of ceftazidime relative to control populations of strain MG (Fig. 2).

We propagated 48 independent populations of each strain (MG, MG1::blaTEM-1 and MG/pBGT) in a lysogeny broth (LB) medium with increasing concentrations of ceftazidime. The experiment started with one-quarter of the MIC of ceftazidime for each parental strain (0.06 mg l−1 for MG and MG::blaTEM-1 and 0.25 mg l−1 for MG/pBGT) and 1% of each population was transferred to fresh medium with the concentration of ceftazidime doubled every day for the duration of the experiment. We measured the number of surviving populations over time in all of the treatments and for MG/pBGT we also measured the plasmid stability and copy number in a subset of the populations. In parallel, we propagated 16 control populations of each strain in LB medium with no antibiotics.

The parental MG strain showed a steep decrease in the number of viable populations after day 4, and all populations of this strain went extinct by day 8 of the experiment (8 mg l−1 ceftazidime). The populations of the strain carrying blaTEM-1 in the chromosome followed the same trajectory as the wild-type MG controls, demonstrating that carrying a single copy of TEM-1 has little, if any, impact on the ability of populations to evolve elevated cephalosporin resistance (viable populations over concentrations, log-rank test, P = 0.373, χ 2 = 0.8, d.f. = 1; Fig. 2). On the other hand, populations of MG/pBGT showed a very different trend compared with the control (viable populations over concentrations, log-rank test, P < 0.001, χ 2 = 41.7, d.f. = 2). Although there was also an important initial decrease in surviving populations, a subset of seven populations was able to survive up to a very high dose of ceftazidime: >4,000 mg l−1 (Fig. 2). This result clearly demonstrates that carrying blaTEM-1 on a multicopy plasmid potentiates the ability of E. coli populations to evolve clinically relevant levels of cephalosporin resistance.

### Molecular basis of ceftazidime resistance in MG/pBGT

We used deep DNA sequencing from samples of whole populations to investigate the genetic basis of ceftazidime resistance. Specifically, we sequenced populations on the last day that at least 50% of the populations were still viable (day 4 for MG/pBGT and day 5 for MG and MG::blaTEM-1, six random populations per strain). Our rationale for this approach is that it is only possible to make meaningful comparisons across treatments by picking a common time point where all of the treatments can be adequately sampled.

Sequencing results from these populations revealed differences in the mutation profiles of MG/pBGT compared with MG::blaTEM-1 and wild-type MG. Populations carrying blaTEM-1 on the chromosome presented a very similar mutation profile to the wild-type strain MG, with mutations in chromosomal genes previously related to ceftazidime resistance but no mutations in blaTEM-1 (Supplementary Data 1). Changes on chromosomal genes included non-synonymous mutations and indels in the transcriptional repressors of active drug efflux systems such as marR, acrR and baeRS, and also in the regulator system envZ-ompR, which activates the expression of the outer-membrane porin proteins OmpF and OmpC33,​34,​35. Populations belonging to the plasmid-bearing MG/pBGT strain carried some mutations similar to the ones described above, such as non-synonymous mutations in envZ-ompR or in ompF. However, unlike MG::blaTEM-1, four out of six plasmid-carrying populations also carried mutations in blaTEM-1. Mutations in TEM-1 were located in residues 164 (R164S, R164C, R164H), 166 (E166K) and 179 (D179G), which have previously been reported to be responsible for ceftazidime resistance in vivo and in vitro 32,36,37. The presence of TEM-1 mutations is consistent with the idea that plasmids accelerate evolution by increasing the supply of mutations in functionally important plasmid-carried genes.

In order to understand the specific mechanisms that drive high-level resistance in MG/pBGT populations, we sequenced a time series of samples from the seven populations that survived up to 4,096 mg l−1 of ceftazidime (days 0, 4, 8, 12 and 16; Supplementary Data 1). Selective sweeps of blaTEM-1 mutations occurred in all seven populations, demonstrating a link between modification in the β-lactamase and high-level ceftazidime resistance (Fig. 3c). Moreover, TEM-1 mutations were already present at detectable frequencies in these populations at day 4, suggesting that early TEM-1 mutations played a key role in determining evolvability in our experiment. Two specific mutations in TEM-1 were present in the seven populations that reached high-level ceftazidime resistance: R164S (5/7 populations) or R164H (2/7 populations)32 (Fig. 3c).

To help differentiate between mutations that are associated with increased resistance, as opposed to general adaptation to the laboratory or hypermutable regions, we also sequenced six control populations of each strain that were evolved for 16 days in the LB medium with no antibiotics. None of the six control populations propagated in LB for 16 days showed mutations in any of the chromosomal genes associated with ceftazidime resistance or in blaTEM-1 (Supplementary Data 1).

### Increase in plasmid copy number is associated with mutations in the RNAI of pBGT

Gene amplification has been demonstrated to play an important role in bacterial evolution, including anti­biotic resistance38,39. To test the role of gene amplification in evolution, we measured the copy number of pBGT through time. The evolution of high-level ceftazidime resistance was associated with an increase in pBGT copy number and in extreme cases the copy number increased approximately tenfold (Fig. 3a). Analysis of the pBGT sequences revealed that the increase in the copy number was associated with the rise of mutations near the origin of replication (oriV) of the plasmid at days 12 and 16 (Fig. 3). These mutations were located in a strategic region that codes for the small RNAs that are responsible for the control of the replication and copy number of ColE1 plasmids: RNAI and RNAII40. Briefly, RNAII acts as the primer precursor for plasmid replication and RNAI binds to RNAII blocking the initiation of replication. These RNAs are coded in opposite strands of the same DNA region and they present a secondary structure of three consecutive hairpins (Fig. 3b), recognizing each other by complementarity41. The two most frequent mutations in the oriV of pBGT were placed in the loop of the central hairpin of RNAI, which is the first binding site between RNAI and RNAII42. Both mutations produced a G to U change in one of two contiguous residues at positions 54 and 55 in the RNAI molecule (Fig. 3b). Interestingly, G/C to A/U mutations in residues in the loop of the central hairpin of RNAI in ColE1-type plasmids are known to be responsible for a reduction in the binding rate constant between RNAI and RNAII due to a lower strength of the initial recognition between the loops41. This lower affinity between the two RNA molecules leads to an increase in the initiation of replication and a subsequent increase in the plasmid copy number43. The remaining mutations in RNAI were present in the stem region of the three different hairpins (Fig. 3). These mutations probably affected the secondary structure of RNAI, leading to a de-repression of plasmid replication43.

### Reconstruction of ceftazidime resistance and high plasmid copy number mutations

Our sequencing results suggested that plasmid carriage of TEM-1 increases the ability of populations to evolve high levels of ceftazidime resistance by facilitating the evolution of novel TEM-1 variants that are capable of hydrolysing ceftazidime more effectively. To test this idea, we reconstructed the most commonly observed TEM-1 substitution (R164S) into both the MG::blaTEM-1 and the MG/pBGT ancestral strains (Fig. 4a). Inserting this mutation into the chromosomal copy of blaTEM-1 resulted in a modest twofold increase in ceftazidime resistance compared with MG::blaTEM-1 (and with MG; Table 1). In contrast, adding mutation R164S into the pBGT plasmid increased ceftazidime resistance 128-fold compared with MG/pBGT (MIC = 64 mg l−1). Importantly, this increase in resistance easily surpasses the clinical breakpoint for ceftazidime resistance (8 mg l−1)44 (Fig. 4b, Table 1). This difference in MIC between plasmid and chromosomal location of the β-lactamase mutation is due to a gene dosage effect previously observed for these enzymes28,29.

Pleiotropic costs could potentially reduce the fitness benefit associated with resistance mutations in high-copy-number plasmids. For example, R164S substitutions in TEM-1 decrease ampicillin resistance as a result of antagonistic pleiotropiy (Table 1, Supplementary Fig. 1)45. However, we did not find any fitness cost in the absence of antibiotics associated with introducing substitution R164S into either the chromosomal (two-sample t-test, P = 0.654, t = −0.47, d.f. = 5.9) or plasmid-carried TEM-1 (two-sample t-test, P = 0.282, t = −1.12, d.f. = 14.9) (Fig. 4d, Table 1). Collectively, these results demonstrate that plasmid pBGT accelerates the evolution of ceftazidime resistance by massively amplifying the increase in resistance associated with TEM-1 mutations. Importantly, this effect is sufficient to explain how plasmid carriage allows populations to evolve levels of ceftazidime resistance that are above the clinical breakpoint.

A second key insight of our sequencing results is that high-level ceftazidime resistance repeatedly evolves by a process of TEM-1 modification followed by an increase in the copy number of pBGT associated to mutations in the RNAI. To check whether the fitness benefits associated with an increased plasmid copy number were dependent on the previous modification of TEM-1, we reconstructed the two most common mutations in RNAI (G54U and G55U) into strains MG/pBGT and MG/pBGT R164S (Fig. 4a). Introducing mutations into RNAI led to large increases in the plasmid copy number, ranging from 40 to 130 copies per cell, confirming the role of these mutations in increased pBGT copy number (Fig. 4c, Table 1). In contrast to the massive increases in resistance associated with TEM-1 substitutions, the increased copy number had subtle effects on ceftazidime resistance, leading to only twofold increases (or less) in MIC (Fig. 4b, Table 1). The increased copy number was associated with profound fitness costs in the absence of antibiotics (Fig. 4d, Table 1); fitness declined at a rate of approximately 0.4% per plasmid copy (Fig. 4e, Pearson’s test, t = −8.56, d.f. = 4, P = 0.001, R 2 = 0.95). From these results it is clear that there is little, if any, impact of TEM-1 mutation on the fitness benefits associated with pBGT amplification. Instead, our results support the idea that TEM-1 mutations preceded pBGT amplification simply because TEM-1 mutations resulted in large increases in resistance without imposing any additional fitness burden on the cell. We argue that the very high cost associated with the elevated plasmid copy number mutants will probably reduce the likelihood of finding these genotypes in nature.

## Discussion

Our results demonstrate that multicopy plasmids provide an evolutionary advantage above and beyond mediating HGT. In the short term, carrying blaTEM-1 on a multicopy plasmid (pBGT) is associated with an increase in fitness under conditions of strong selection for β-lactam resistance. In the long term, plasmid pBGT acts as an evolutionary catalyst that facilitates the evolution of novel variants of blaTEM-1 and allows bacterial populations to evolve clinically relevant levels of ceftazidime resistance. Because pBGT is a multicopy plasmid, carrying blaTEM-1 on the plasmid ensures an increased rate of supply of beneficial mutations in blaTEM-1 with an improved rate of ceftazidime hydrolysis. The phenotypic effect of these mutations is then amplified by increased expression arising from the constitutive high copy number of the pBGT plasmid (19 copies per cell). This step of amplification of the mutation due to the multicopy nature of the plasmid is crucial, as shown by the fact that the same blaTEM-1 mutation that increases the MIC of ceftazidime over the clinical breakpoint when encoded in the plasmid, has very little effect when cloned as a single copy in the chromosome (Fig. 4).

Gene amplification and modification are thought to play key roles in the evolution of novel bacterial phenotypes38,39,46,​47,​48. We argue that pBGT is such an effective evolutionary catalyst because it provides a vehicle for the simultaneous modification of blaTEM-1 and expansion in the frequency of the mutated allele. This platform provides an enhanced ‘gene duplication-amplification’ system for adaptive evolution48, which aligns with the ‘amplification mutagenesis’ model proposed by Hendrickson and colleagues19. Although it is easy to understand how multicopy plasmids accelerate evolution by increased mutation supply using classical population genetics, it is more challenging to understand the amplification of plasmid-carried mutations. An important consequence of the partitioning of small plasmids at cell division is that copies of plasmids carrying beneficial mutations are randomly distributed between daughter cells. When combined with plasmid replication, this random partitioning ensures that a substantial fraction of the daughter cells have an increased number of copies of beneficial mutations relative to their mother cells (Supplementary Fig. 2). Crucially, this amplification of beneficial mutations by random segregation is expected to occur at a much greater rate than by the spontaneous duplication of genes carrying beneficial mutations. Selection will subsequently fuel the spread of the cells that carry a higher frequency of plasmid-born mutated alleles, driving the rapid increase in frequency of the mutation at both the cellular and population level in a few generations (Supplementary Fig. 2). However, the dynamics of this process are potentially very complex, and we are currently developing deterministic population dynamics models and stochastic agent-based simulations to better understand this process.

One important feature of our experimental design is that we used a rapid ramp-up of antibiotics, which may have reduced the accessibility to alternative evolutionary trajectories to high-level β-lactam resistance49, and this may have maximized the evolutionary benefit conferred by pBGT. However, we argue that our experimental design is relevant in the context of evolution of antibiotic resistance because sharp temporal and spatial gradients in antibiotic concentrations are common in patients50. Another limitation of this study is the fact that pBGT carries three extra genes apart from blaTEM-1 (Fig. 1). However, pBGT mutations were located in blaTEM-1 or the RNAI, so there is no evidence indicating a possible role for the extra genes on the evolution of ceftazidime resistance.

In this work we used TEM-1 as model system, but we argue that our results have general implications for the evolution of antibiotic resistance, adaptation and innovation in bacteria. Small multicopy plasmids are very common in pathogenic bacteria12,28 and they frequently carry antibiotic resistance genes20,​21,​22,26, but their role in the ecology and evolution of antibiotic resistance has generally been overlooked. Our work shows that these genetic elements could accelerate the evolution of resistance in vivo. Further work will be required to test this possibility.

## Methods

### Culture conditions and antibiotic susceptibility testing

The bacterial strains, plasmids and primers used in this study are listed in Supplementary Tables 1–3. Bacterial strains were cultured in LB broth at 37 °C in 96-well plates with continuous shaking (225 r.p.m.) and on LB agar plates at 37 °C (Fisher Scientific). Minimal inhibitory concentrations (MIC) were determined in LB (as the experimental evolution assay) following Clinical and Laboratory Standards Institute guidelines51. To determine the different MICs, we performed six biological replicates for each strain, and we used the mode of the results as the MIC value.

### Construction of the experimental system

The E. coli MG1655 strain was used for all of our constructions. Chromosomal insertion of blaTEM-1 was designed to take place in the integration site of the λ phage (attB)52 as follows: the blaTEM-1 gene was amplified from plasmid p393853 with primers 3658 and 3659. We PCR-amplified the regions at both sides of the chromosomal attB site (primers 3652–3653 and 3654–3655) and fused them to the blaTEM-1 PCR product using Gibson Assembly54. We electroporated the fusion of these fragments into an MG1655 strain containing the pKOBEG plasmid55. This thermo-sensitive plasmid contains the λ Red machinery, allowing the production of homology-based allelic exchanges between the chromosome and PCR products56. Selection for carbenicillin resistance at 37 °C yielded clones in which blaTEM-1 had been integrated in attB and pKOBEG was lost. Verification of blaTEM-1 sequence and pKOBEG loss gave rise to strain MG::blaTEM-1. The allele of blaTEM-1 coding for R164S mutation was introduced in the non-evolved strain using the same methodology.

To assess the evolvability of genes when located on plasmids, we cloned blaTEM-1 in p3655 (pSU18T-pBADgfp2, ColE1-type origin of replication57) producing plasmid pBGT. To do so, we phosphorylated the PCR product of primers 4304–4305 (containing blaTEM-1), and ligated it to the PCR-amplified backbone of p3655 (primers 3086–738). Ligations were transformed into E. coli DH5-α and selected on carbenicillin and chloramphenicol then tested for gfp induction with l-arabinose. Sequence of blaTEM-1 was verified. pBGT was transformed into the parental MG1655 producing the strain MG/pBGT.

The combination of evolved replication origins with wild-type blaTEM-1 alleles in pBGT was performed using Gibson assembly. Briefly, replication origins were amplified using primers 837 and 296 and assembled to the backbone of pBGT amplified with primers 737 and 1701. Verification of the sequence of the oriV of the plasmids together with the selection on carbenicillin, chloramphenicol and l-arabinose (to induce GFP expression) allowed confirmation of the functionality of all plasmid components.

### Competitive fitness assays

The fitness of each clone was determined basically as described in San Millan and colleagues9. We performed competitions between a plasmid-bearing clone and a plasmid-free clone in each case (MG/pBGT and MG::blaTEM-1 were the control strains to compete against). Four biological replicates of the competition (of four technical replicates each) were performed for each clone. Pre-cultures of the clones were incubated at 37 °C with 225 r.p.m. shaking overnight in 96-well plates carrying 200 μl of LB broth per well (Fisher Scientific). Pre-cultures were diluted 400-fold in 200 μl of fresh LB and mixed at a ratio of approximately 50% of each clone. The exact initial proportions were confirmed via flow cytometry using an Accuri C6 Flow Cytometer Instrument (BD Accuri, San Jose) with the following parameters: flow rate: 66 μl min−1, core size: 22 μm, events recorded per sample: 10,000. To measure these proportions we incubated an aliquot of the mix in LB with l-arabinose 0.25% for 2 h to induce GFP expression from pBGT. Mixtures were competed for 24 h in LB at 37 °C with 225 r.p.m. shaking (approximately eight generations). Again, the final proportion was measured by flow cytometry as described above. The fitness of the strain carrying the plasmid relative to the plasmid-free strain was determined using the formula: $W p + = ln ( N final , p + N initial , p + ) ln ( N final , p − N initial , p − )$ where Wp+ is the relative fitness of the plasmid-bearing clone, Ninitial,p+ and Nfinal,p+ are the numbers of cells of the plasmid-carrying clone before and after the competition, and Ninitial,p- and Nfinal,p- are the numbers of cells of the pBGT-free clone before and after the competition. As a control, MG/pBGT and MG::blaTEM-1 were competed in every experiment.

### Experimental evolution

We cultured strains (MG, MG/pBGT and MG::blaTEM-1) on LB agar plates at 37 °C to obtain isolated colonies. We inoculated single colonies in alternative wells of 96-well plates containing LB broth (48 populations per plate, one plate per strain). We used this checkerboard plate design of intercalating inoculated wells with bacteria-free medium to prevent and control for cross-contamination in the plates. We started the experimental evolution assay inoculating 2 μl of the initial overnight cultures into 198 μl of LB with one quarter of the MIC of ceftazidime of each strain. Every day we transferred 2 μl of the overnight culture into 198 μl of fresh medium with double the concentration of ceftazidime of the day before (approximately 6–7 generations per day). In parallel, we propagated 16 control populations of each strain in the same conditions as previously described but in the absence of ceftazidime. Populations with positive growth were defined by an optical density value (measured overnight at a wavelength of 600 nm) higher than 0.1. We kept a frozen stock of all of the populations from day 0, 4, 8, 12 and 16. On those days we also checked the plasmid stability by flow cytometry (using l-arabinose induction, as explained in the previous section) and performed DNA extraction to quantify plasmid copy number in the population (see section below), from four populations belonging to MG/pBGT. We also measured the plasmid stability every day from 10 of the populations of MG/pBGT evolving in the absence of antibiotics.

### Quantification of plasmid copy number

The copy number of pBGT was determined by quantitative polymerase chain reaction (qPCR) using an ABI StepOnePlus Real-Time PCR System (Life Technologies). DNA extraction, quantification and digestion (with BamHI-HF, New England Biolabs) were performed as previously described27. We developed a specific qPCR for pBGT (pBGT-F: ACATTTCCGTGTCGCCCTT, pBGT-R: CACTCGTGCACCCAACTGA, amplicon size: 115 bp, efficiency: 94.29%, R 2 = 0.99) and we used a previously described qPCR for the dxs chromosomal monocopy gene (dxs-F: CGAGAAACTGGCGATCCTTA, dxs-R: CTTCATCAAGCGGTTTCACA, amplicon size: 113 bp, efficiency: 95.52%, R 2 = 0.99)58 to compare the ratio of plasmid and chromosomal DNA. The efficiency of the reactions was calculated from the standard curve generated by performing qPCR with four eightfold dilutions of template DNAs in triplicate (approximately 5 ng μl−1 to 12 pg μl−1 working range of DNA concentration). qPCRs were performed using ABI SYBR Select Master Mix (Life Technologies) at a final DNA concentration of 0.1 ng μl−1 and following manufacturers’ instructions. The amplification conditions were: initial denaturation for 2 min at 95 °C, followed by 40 cycles of denaturation for 15 s at 95 °C, annealing and extension for 1 min at 60 °C. After the amplification was complete and to control for the specificity of the reaction, a melting curve analysis was performed by cooling the reaction to 60 °C and then heating slowly to 95 °C. Inter-run calibration samples were used to normalize the results from different plates of each qPCR. The copy number was calculated using the ΔΔCT method as previously described, given that the amplification efficiencies of the target and reference genes were approximately equal58. The average plasmid copy number was corrected using the plasmid stability frequencies, which were measured from the cultures used to obtain the DNA samples.

### Genome sequencing and bioinformatic analysis

We sequenced six random populations from each strain from the experimental evolution assay on the last day when at least 50% of the populations were still viable. We also sequenced a time series of samples from the seven MG/pBGT populations that survived up to 4,096 mg l−1 of ceftazidime (days 0, 4, 8, 12 and 16) and nine MG/pBGT clones from three populations with increased plasmid copy number. Finally, we also sequenced six control populations of each strain evolved for 16 days in LB medium with no antibiotics (Supplementary Data 1). DNA extractions were performed from 2 ml of LB broth (Fisher Scientific) cultures incubated at 37 °C with 225 r.p.m. shaking overnight using the Promega Wizard Genomic 4 DNA Purification Kit (Promega). DNA was quantified using the QuantiFluor dsDNA system (Promega), following manufacturers’ instructions. All sequencing was conducted at the Wellcome Trust Centre for Human Genetics using the Illumina HiSeq2500 platform. Mutations were predicted using the breseq 0.26.1 pipeline59,60, using polymorphism mode to estimate the frequency of mutations in populations. Mutations classes included point mutations, small (<50 bp) and large (>50 bp) indels, copy number variation and insertion-sequence-element insertions61. Reads were aligned to the E. coli MG1655 reference genome (NC_000913.3), the TEM-1 chromosomal insertion (this study), and the pBGT plasmid genome (this study). Variants that never surpassed 10% frequency in any population were filtered from the dataset. Mutations that were present in the ancestral clones relative to the reference sequence were also excluded, leaving only mutations that accumulated throughout the experiment. We sequenced the DNA from the three parental strains and from a total of 51 DNA samples obtained from evolving populations (Supplementary Data 1). In addition, we sequenced DNA samples from clones of plasmid-bearing populations showing especially high plasmid copy numbers (three clones/populations from three different populations: nine clones in total; Supplementary Data 1), to successfully confirm the presence of the mutations observed in the populations. Two out of the 63 samples were discarded due to results showing possible DNA contamination: one sample from MG and the other from MG::blaTEM-1, both of them from populations at day 5.

The structure of the RNAI of plasmid pBGT was predicted using the default parameters of RNAfold software62. The result was compared with previous reports of the structure of this molecule41,43 from oriV of almost identical ColE1 plasmids, confirming the homology.

### Statistical analysis

Analyses were performed using RStudio (Version 0.99.486). Comparisons among strains were performed using two-sample t-tests. Survival of populations over time under increasing concentration of ceftazidime was analysed using log-rank tests. The variation of the plasmid copy number in MG/pBGT populations over time was modelled using a generalized linear effects model, with time and treatments (and their interaction) as factors and assuming Gaussian error distribution.

### Data availability

The sequence data that support the findings of this study have been deposited in the European Nucleotide Archive with the accession code PRJEB14587 (http://www.ebi.ac.uk/ena/data/view/PRJEB14587).

How to cite this article: San Millan, A., Escudero, J. A., Gifford, D. R., Mazel, D. & MacLean, R. C. Multicopy plasmids potentiate the evolution of antibiotic resistance in bacteria. Nat. Ecol. Evol. 1, 0010 (2016).

## References

1. 1.

, & Lateral gene transfer and the nature of bacterial innovation. Nature 405, 299–304 (2000).

2. 2.

& Origins of bacterial diversity through horizontal genetic transfer and adaptation to new ecological niches. FEMS Microbiol. Rev. 35, 957–976 (2011).

3. 3.

Exploring the costs of horizontal gene transfer. Trends Ecol. Evol. 28, 489–495 (2013).

4. 4.

& The genetic basis of the fitness costs of antimicrobial resistance: a meta-analysis approach. Evol. Appl. 8, 284–295 (2014).

5. 5.

& The population biology of bacterial plasmids: a priori conditions for the existence of conjugationally transmitted factors. Genetics 87, 209–228 (1977).

6. 6.

& The population biology of bacterial plasmids: a priori conditions for the existence of mobilizable nonconjugative factors. Genetics 94, 425–443 (1980).

7. 7.

, & Natural selection, infectious transfer and the existence conditions for bacterial plasmids. Genetics 155, 1505–1519 (2000).

8. 8.

& Plasmid-mediated horizontal gene transfer is a coevolutionary process. Trends Microbiol. 20, 262–267 (2012).

9. 9.

et al. Positive selection and compensatory adaptation interact to stabilize non-transmissible plasmids. Nat. Commun. 5, 5208 (2014).

10. 10.

, , , & Parallel compensatory evolution stabilizes plasmids across the parasitism–mutualism continuum. Curr. Biol. 25, 2034–2039 (2015).

11. 11.

, , & Evaluating the effect of horizontal transmission on the stability of plasmids under different selection regimes. Mob. Genet. Elements 22, 1–5 (2015).

12. 12.

, , , & Genetic changes accompanying increased fitness in evolving populations of Escherichia coli . Genetics 130, 241–249 (1992).

13. 13.

et al. Evolutionary history of the global emergence of the Escherichia coli epidemic clone ST131. MBio 7, e02162 (2016).

14. 14.

, , & Chromosomal stasis versus plasmid plasticity in aphid endosymbiont Buchnera aphidicola . Heredity 95, 339–347 (2005).

15. 15.

, & Positive selection for elevated gene expression noise in yeast. Mol. Syst. Biol. 5, 299 (2009).

16. 16.

& Mutation frequencies and antibiotic resistance. Antimicrob. Agents Chemother. 44, 1771–1777 (2000).

17. 17.

, & Bypass of genetic constraints during mutator evolution to antibiotic resistance. Proc. Biol. Sci. 282, 20142698 (2015).

18. 18.

, , , & Plasmid copy number underlies adaptive mutability in bacteria. Genetics 198, 919–933 (2014).

19. 19.

, , , & Amplification-mutagenesis: evidence that "directed" adaptive mutation and general hypermutability result from growth with a selected gene amplification. Proc. Natl Acad. Sci. USA 99, 2164–2169 (2002).

20. 20.

, , , & Complete nucleotide sequence of Klebsiella pneumoniae multiresistance plasmid pJHCMW1. Antimicrob. Agents Chemother. 46, 3422–3427 (2002).

21. 21.

et al. Molecular organization of small plasmids bearing blaTEM-1 and conferring resistance to beta-lactams in Haemophilus influenzae . Antimicrob. Agents Chemother. 56, 4958–4960 (2012).

22. 22.

et al. New TEM-derived extended-spectrum beta-lactamase and its genomic context in plasmids from Salmonella enterica serovar derby isolates from Uruguay. Antimicrob. Agents Chemother. 50, 781–784 (2006).

23. 23.

, , , & Robustness–epistasis link shapes the fitness landscape of a randomly drifting protein. Nature 444, 929–932 (2006).

24. 24.

, & Evolvability as a function of purifying selection in TEM-1 beta-lactamase. Cell 160, 882–892 (2015).

25. 25.

, , & Quantifying the adaptive potential of an antibiotic resistance enzyme. PLoS Genet. 8, e1002783 (2012).

26. 26.

et al. Small-plasmid-mediated antibiotic resistance is enhanced by increases in plasmid copy number and bacterial fitness. Antimicrob. Agents Chemother. 59, 3335–3341 (2015).

27. 27.

, & Positive epistasis between co-infecting plasmids promotes plasmid survival in bacterial populations. ISME J. 8, 601–612 (2014).

28. 28.

, , & Synergistic effect of dosage and bacterial inoculum in TEM-1 mediated antibiotic resistance. Eur. J. Clin. Microbiol. Infect. Dis. 7, 778–779 (1988).

29. 29.

et al. Resistance to beta-lactam/clavulanate. Lancet 2, 1473 (1987).

30. 30.

, , , & Isolated cell behavior drives the evolution of antibiotic resistance. Mol. Syst. Biol. 11, 822 (2015).

31. 31.

, , , & Selection of a multidrug resistance plasmid by sublethal levels of antibiotics and heavy metals. MBio 5, e01918–14 (2014).

32. 32.

, & Natural evolution of TEM-1 β-lactamase: experimental reconstruction and clinical relevance. FEMS Microbiol. Rev. 34, 1015–1036 (2010).

33. 33.

, , & Resistance to ceftazidime in Escherichia coli associated with AcrR, MarR and PBP3 mutations and overexpression of sdiA. J. Med. Microbiol. 63, 56–65 (2014).

34. 34.

, , , & Beta-lactam resistance modulated by the overexpression of response regulators of two-component signal transduction systems in Escherichia coli . J. Antimicrob. Chemother. 52, 576–582 (2003).

35. 35.

& EnvZ-OmpR interaction and osmoregulation in Escherichia coli . J. Biol. Chem. 277, 24155–24161 (2002).

36. 36.

et al. In vivo selection of a complex mutant TEM (CMT) from an inhibitor-resistant TEM (IRT) during ceftazidime therapy. J. Antimicrob. Chemother. 68, 2792–2796 (2013).

37. 37.

, , , & Concentration-dependent selection of small phenotypic differences in TEM beta-lactamase-mediated antibiotic resistance. Antimicrob. Agents Chemother. 44, 2485–2491 (2000).

38. 38.

, , & Real-time evolution of new genes by innovation, amplification, and divergence. Science 338, 384–387 (2012).

39. 39.

& Bacterial gene amplification: implications for the evolution of antibiotic resistance. Nat. Rev. Microbiol. 7, 578–588 (2009).

40. 40.

, , & Inhibition of ColE1 RNA primer formation by a plasmid-specified small RNA. Proc. Natl Acad. Sci. USA 78, 1421–1425 (1981).

41. 41.

Control of ColE1 plasmid replication: the process of binding of RNA I to the primer transcript. Cell 38, 861–870 (1984).

42. 42.

& Interaction between RNA1 and the primer precursor in the regulation of Co1E1 replication. J. Mol. Biol. 170, 635–650 (1983).

43. 43.

& Base pairing of RNA I with its complementary sequence in the primer precursor inhibits ColE1 replication. Nature 294, 623–626 (1981).

44. 44.

Breakpoint Tables for Interpretation of MICs and Zone Diameters v. 6.0. (The European Committee on Antimicrobial Susceptibility Testing, 2016);

45. 45.

, & Evolution of an antibiotic resistance enzyme constrained by stability and activity trade-offs. J. Mol. Biol. 320, 85–95 (2002).

46. 46.

Evolution by Gene Duplication (Springer, 1970).

47. 47.

, , & The genomic basis of evolutionary innovation in Pseudomonas aeruginosa. PLoS Genet. 12, e1006005 (2016).

48. 48.

& Gene amplification and adaptive evolution in bacteria. Annu. Rev. Genet. 43, 167–195 (2009).

49. 49.

, , & Evolutionary rescue from extinction is contingent on a lower rate of environmental change. Nature 494, 463–467 (2013).

50. 50.

& Selective compartments for resistant microorganisms in antibiotic gradients. Bioessays 19, 731–736 (1997).

51. 51.

Performance standards for antimicrobial susceptibility testing. 19th ed. (Clinical and Laboratory Standards Institute, 2009).

52. 52.

et al. Unmasking the ancestral activity of integron integrases reveals a smooth evolutionary transition during functional innovation. Nat. Commun. 7, 10937 (2016).

53. 53.

, , & Identification of key structural determinants of the IntI1 integron integrase that influence attC x attI1 recombination efficiency. Nucleic Acids Res. 35, 6475–6489 (2007).

54. 54.

et al. Enzymatic assembly of DNA molecules up to several hundred kilobases. Nat. Methods 6, 343–345 (2009).

55. 55.

, & A rapid method for efficient gene replacement in the filamentous fungus Aspergillus nidulans. Nucleic Acids Res. 28, E97 (2000).

56. 56.

& One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. Proc. Natl Acad. Sci. USA 97, 6640–6645 (2000).

57. 57.

, , & Construction of a Vibrio splendidus mutant lacking the metalloprotease gene vsm by use of a novel counterselectable suicide vector. Appl. Environ. Microbiol. 73, 777–784 (2007).

58. 58.

, , & Absolute and relative QPCR quantification of plasmid copy number in Escherichia coli . J. Biotechnol. 123, 273–280 (2006).

59. 59.

& Identification of mutations in laboratory-evolved microbes from next-generation sequencing data using breseq. Methods Mol. Biol. 1151, 165–1882 (2014).

60. 60.

et al. Identifying structural variation in haploid microbial genomes from short-read resequencing data using breseq. BMC Genomics 15, 1039 (2014).

61. 61.

, , & Detecting rare structural variation in evolving microbial populations from new sequence junctions using breseq. Front. Genet. 5, 468 (2014).

62. 62.

et al. ViennaRNA Package 2.0. Algorithms Mol. Biol. 6, 26 (2011).

## Acknowledgements

A.S.M. thanks E. Frago for his help with the statistical analyses. This work was supported by funding from the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013)/ERC grant (StG-2011-281591). A.S.M. is supported by a Miguel Servet fellowship from the Instituto de Salud Carlos III (MS15/00012) co-financed by the European Social Fund and The European Development Regional Fund ‘A way to achieve Europe’ (ERDF). J.A.E. is supported by a Marie Curie Intra-European Fellowship for Career Development (FP-7-PEOPLE-2011-IEF, ICADIGE). Work in the Mazel lab was supported by the Institut Pasteur, the Centre National de la Recherche Scientifique (CNRS-UMR3525) and the European Union Seventh Framework Programme (FP7-HEALTH- 2011-single-stage), the ‘Evolution and Transfer of Antibiotic Resistance’ (EvoTAR, FP7- HEALTH-282004).

We thank the High-Throughput Genomics Group at the Wellcome Trust Centre for Human Genetics funded by Wellcome Trust grant reference 090532/Z/09/Z and Medical Research Council Hub grant G0900747 91070 for generation of the high-throughput sequencing data.

## Author notes

• Jose Antonio Escudero
•  & Danna R. Gifford

These authors contributed equally to this work.

## Affiliations

1. ### Department of Zoology, University of Oxford, Oxford OX1 3PS, UK

• Alvaro San Millan
• , Danna R. Gifford
•  & R. Craig MacLean
2. ### Department of Microbiology, Hospital Universitario Ramon y Cajal (IRYCIS), 28034 Madrid, Spain

• Alvaro San Millan
3. ### Institut Pasteur, Unité de Plasticité du Génome Bactérien, Département Génomes et Génétique, 28 Rue du Dr. Roux, 75015 Paris, France

• Jose Antonio Escudero
•  & Didier Mazel
4. ### CNRS, UMR3525, 28 Rue du Dr. Roux, 75015 Paris, France

• Jose Antonio Escudero
•  & Didier Mazel

## Authors

### Contributions

A.S.M. and R.C.M. were responsible for the conceptualization of this study; A.S.M., J.A.E., and D.R.G. designed the methodology; formal analysis was by D.R.G.; A.S.M. and J.A.E. carried out the investigations; A.S.M. and R.C.M. prepared the original draft of the manuscript and also undertook the reviewing and editing; R.C.M. and D.M. were responsible for funding acquisition and for supervision; D.M. oversaw the resources; project administration was performed by A.S.M.

### Competing interests

The authors declare no competing financial interests.

## Corresponding author

Correspondence to Alvaro San Millan.

## PDF files

1. 1.

### Supplementary Information

Supplementary Figures 1,2, Supplementary Tables 1–3 and Supplementary References.

## Excel files

1. 1.

### Supplementary Data 1

Description of the mutations affecting the chromosome and pBGT in the different populations and clones of MG, MG::blaTEM-1 and MG/pBGT analysed in this work. The position and nature of the mutations are indicated, as well as the genes affected and their activity. The frequencies of the mutations in the different populations and clones are also indicated.