Amelioration of the Cost of Conjugative Plasmid Carriage in Eschericha coli K 12

Although plasmids can provide beneficial functions to their host bacteria, they might confer a physiological or energetic cost. This study examines how natural selection may reduce the cost of carrying conjugative plasmids with drug-resistance markers in the absence of antibiotic selection. We studied two plasmids, R1 and RP4, both of which carry multiple drug resistance genes and were shown to impose an initial fitness cost on Escherichia coli. To determine if and how the cost could be reduced, we subjected plasmid-containing bacteria to 1100 generations of evolution in batch cultures. Analysis of the evolved populations revealed that plasmid loss never occurred, but that the cost was reduced through genetic changes in both the plasmids and the bacteria. Changes in the plasmids were inferred by the demonstration that evolved plasmids no longer imposed a cost on their hosts when transferred to a plasmid-free clone of the ancestral E. coli. Changes in the bacteria were shown by the lowered cost when the ancestral plasmids were introduced into evolved bacteria that had been cured of their (evolved) plasmids. Additionally, changes in the bacteria were inferred because conjugative transfer rates of evolved R1 plasmids were lower in the evolved host than in the ancestral host. Our results suggest that once a conjugative bacterial plasmid has invaded a bacterial population it will remain even if the original selection is discontinued. OUR capacity to decrease antibiotic resistance among Lenski 1988; Modi and Adams 1991). Since these elements are also transmitted vertically during cell division clinically important bacteria depends in part on the fitness costs associated with the resistances. If a cost is one would anticipate that natural selection operating on the host, on the element, or both would either ameliorate associated with a resistance phenotype, sensitive strains should be able to invade and outcompete resistant bacthe fitness costs or favor the ascent of bacteria that have lost these elements, i.e., segregation (Levin and Lenski terial populations when the use of antibiotics is reduced. However, several studies examining the effects of reduc1983). Amelioration of the cost of carriage for simpler, nonconjugative plasmids has been obtained by host evoluing antibiotic use have shown that for chromosomally encoded resistances the cost is often ameliorated by comtion after evolution with selecting antibiotics (Bouma and Lenski 1988) as well as plasmid-host coevolution when pensatory mutations without loss of resistance (Shrag and Perrot 1996; Shrag et al. 1997; Björkman et al. 1998, evolution occurred without antibiotic selection (Modi and Adams 1991). Although the data are scarce, naturally oc2000; Reynolds 2000; Nagaev et al. 2001). The potential cost to a bacterial host of plasmid-borne curring conjugative resistance plasmids are also generally considered costly when there is no selection for the resisantibiotic resistance may be associated with the resistance function itself or with plasmid regulation. For instance, tance phenotype (Godwin and Slater 1979; Zünd and Lebek 1980). A major difference between the high-copyplasmid-borne antibiotic resistances often act through alteration or efflux of the antibiotic and these mechanisms number cloning vector type of plasmids used in the experiments mentioned above and many of the conjugative lowmay disturb bacterial growth. In addition, plasmid replicacopy-number plasmids is, in addition to their potential tion and gene expression may interfere with bacterial for horizontal dissemination within and between bacterial growth in ways independent of any resistance functions species, the presence of stability systems in the latter. These carried by that plasmid. These costs can be expected to systems are found in most conjugative plasmids and ensure be specific for each plasmid/bacteria combination. faithful inheritance to the daughter cells at cell division. In the absence of the selecting antibiotics, a cost associSince plasmids contribute greatly to antibiotic-resistance ated with plasmid carriage has been observed for noncondevelopment in bacteria, it is fundamentally important to jugative plasmids bearing antibiotic-resistance genes understand how plasmid-borne resistances may respond (Zünd and Lebek 1980; Helling et al. 1981; Noack et al. to a decreased use of antibiotics. This report focuses on 1981; Caulcott et al. 1983; Dykhuizen and Hartl 1983; two well-characterized plasmids with different replication Cooper et al. 1987; Lenski and Bouma 1987; Bouma and systems, R1 ( 100 kb) and RP4 (60 kb). We examine how the costs of these plasmids on an Escherichia coli host can be ameliorated by natural selection in the absence of 1Corresponding author: Swedish Institute for Infectious Disease Control, Solna, S 171 82, Sweden. E-mail: cecilia.dahlberg@smi.ki.se antibiotics. Genetics 165: 1641–1649 (December 2003) 1642 C. Dahlberg and L. Chao

O UR capacity to decrease antibiotic resistance among Lenski 1988; Modi and Adams 1991).Since these ele- ments are also transmitted vertically during cell division clinically important bacteria depends in part on the fitness costs associated with the resistances.If a cost is one would anticipate that natural selection operating on the host, on the element, or both would either ameliorate associated with a resistance phenotype, sensitive strains should be able to invade and outcompete resistant bac-the fitness costs or favor the ascent of bacteria that have lost these elements, i.e., segregation (Levin and Lenski terial populations when the use of antibiotics is reduced.However, several studies examining the effects of reduc-1983).Amelioration of the cost of carriage for simpler, nonconjugative plasmids has been obtained by host evolu-ing antibiotic use have shown that for chromosomally encoded resistances the cost is often ameliorated by com-tion after evolution with selecting antibiotics (Bouma and Lenski 1988) as well as plasmid-host coevolution when pensatory mutations without loss of resistance (Shrag and Perrot 1996;Shrag et al. 1997;Bjo ¨rkman et al. 1998, evolution occurred without antibiotic selection (Modi and Adams 1991).Although the data are scarce, naturally oc-2000;Reynolds 2000;Nagaev et al. 2001).
The potential cost to a bacterial host of plasmid-borne curring conjugative resistance plasmids are also generally considered costly when there is no selection for the resis-antibiotic resistance may be associated with the resistance function itself or with plasmid regulation.For instance, tance phenotype (Godwin and Slater 1979;Zu ¨nd and Lebek 1980).A major difference between the high-copy-plasmid-borne antibiotic resistances often act through alteration or efflux of the antibiotic and these mechanisms number cloning vector type of plasmids used in the experiments mentioned above and many of the conjugative low-may disturb bacterial growth.In addition, plasmid replicacopy-number plasmids is, in addition to their potential tion and gene expression may interfere with bacterial for horizontal dissemination within and between bacterial growth in ways independent of any resistance functions species, the presence of stability systems in the latter.These carried by that plasmid.These costs can be expected to systems are found in most conjugative plasmids and ensure be specific for each plasmid/bacteria combination.
faithful inheritance to the daughter cells at cell division.In the absence of the selecting antibiotics, a cost associ-Since plasmids contribute greatly to antibiotic-resistance ated with plasmid carriage has been observed for noncondevelopment in bacteria, it is fundamentally important to jugative plasmids bearing antibiotic-resistance genes understand how plasmid-borne resistances may respond (Zu ¨nd and Lebek 1980; Helling et al. 1981;Noack et al. to a decreased use of antibiotics.This report focuses on 1981; Caulcott et al. 1983;Dykhuizen and Hartl 1983; two well-characterized plasmids with different replication Cooper et al. 1987;Lenski and Bouma 1987;Bouma and systems, R1 ‫001ف(‬ kb) and RP4 (60 kb).We examine how the costs of these plasmids on an Escherichia coli host can be ameliorated by natural selection in the absence of 1 Corresponding author: Swedish Institute for Infectious Disease Control, Solna, S 171 82, Sweden.E-mail: cecilia.dahlberg@smi.ki.se antibiotics.

MATERIALS AND METHODS
supplemented with 230 g/ml proline, 45 g/ml methionine, and with a glucose concentration of 100 g/ml for liquid cul-Strains and designations: Table 1 lists the E. coli K12 strains tures.This glucose concentration gave a stationary-phase bacteand plasmids used in this study.The plasmids R1 and RP4 were rial concentration of ‫5ف‬ ϫ 10 8 cells/ml.DM agar with a glucose selected for this study on the basis of their differences in replicaconcentration of 1000 g/ml was used for plating.Final concention control and host range.Plasmid R1 has an inhibitor-target trations of antibiotics in Luria agar were 100 g/ml nalidixic mechanism of replication control and has a narrow host range acid, 25 g/ml kanamycin sulfate, 10 g/ml tetracycline, 25 whereas RP4 has an iteron-based control of replication and a g/ml chloramphenicol, 20 g/ml streptomycin, and 100 g/ very broad host range.A notation where b refers to the bacterial ml ampicillin.For phage resistance selection, a lysate of phage host and p to the plasmid it carries was used for this study.
T5 corresponding to 10 9 phages was spread onto DM plates prior The clones used to found long-term serial cultures were hence to spreading of the bacteria.designated b 0 p 0 or b 0 and clones isolated at the end of the experi-Long-term serial cultures: A schematic presentation of the ment were designated b 1 p 1 or b 1 * (b 1 * was used for clones that experimental evolution and the construction of new plasmidhad evolved in the absence of plasmids to distinguish them from host combinations are shown in Figure 1.Three replicate populaconstructions where b 1 p 1 were turned into b 1 by curing).Further tions 1, 2, and 3 of CD104 and three replicate populations 4, 5, strain constructions made from these isolates were designated and 6 of CD105 were initiated from stationary-phase cultures as shown in Figure 1  designations), the bacteria and plasmid used to start either of was, respectively, introduced by conjugation.Strain CD100 is a these triplicate populations were all designated b 0 p 0 .Population spontaneous mutant of J53-1 resistant to bacteriophage T5 and 7 was started with the plasmid-free J53-1 and the starting clone was used as a common competitor in all the pairwise competition was given the designation b 0 .The populations were serially propagated every 24 hr with 40 l of culture into 4 ml of fresh DM experiments (see below).
Growth media: The bacterial populations were propagated without selecting antibiotics in 16-mm tubes.The cultures were incubated at 37Њ in a slanted rack, shaking at 250 rpm.Aerosol-in Davis minimal (DM) medium (Carlton and Brown 1981) lactoside (IPTG)-containing plates.The addition of IPTG results in a large production of CopA RNA that will inhibit R1 replication and make the isolation of evolved R1 segregants possible.After confirmation that the evolved R1 plasmid had been lost the clones were grown in the absence of selecting antibiotics and natural segregants of pKG339 were isolated [CD120-122 (b 1 )].For clones CD114-116 the mini-RP4 replicon pFF1ts97 (Fang and Helinski 1991;Valla et al. 1991) was used for curing.This temperature-sensitive replicon was transformed into the cells and transformants were grown at 30Њ.The pFF1ts97 replicon is incompatible with the wild-type RP4 plasmid and selection for pFF1ts97 made it possible to select for segregants of the evolved RP4 plasmids.After one additional plating and confirmation that the evolved RP4 plasmid was lost, the clones were allowed to grow without selective antibiotics at 37Њ, which would lead to segregation of pFF1ts97, and clones CD123-124 (b 1 ) were then isolated.Each step in the curing procedures was monitored by electrophoresis of plasmid isolations.Controls were made where the plasmids that had been removed by either of the curing Figure 1.-Experimental evolution and strain construction.
procedures above were reintroduced into their host.The ancesb refers to the bacterial host and p to the plasmid it carries.
tral plasmids were finally introduced into the evolved, cured (A) Three independent populations of the E. coli host J53-1 clones and CD117, resulting in CD125-129 (b 1 p 0 ) and CD148carrying either plasmid R1 or plasmid RP4 (b 0 p 0 ) were propa-149 (b 1 *p 0 ).gated by daily serial transfer.Clones were isolated from each Estimates of fitness: Pairwise competition experiments were population after 166 days (b 1 p 1 ).One random clone from each performed against a common competitor, CD100.The competpopulation was subjected to curing, leading to a plasmid-free ing clones were separately preconditioned in DM for one 24-hr evolved host (b 1 ).Curing was accomplished through infection cycle.For each competition experiment cultures were inoculated by an incompatible replicon: pKG339 for R1-carrying clones with the preconditioned cultures of the competitors at a 1:1 and pFF1ts97 for RP4-carrying clones (see main text for deratio and were then propagated for one 24-hr cycle as described tails).The respective ancestral plasmid (p 0 ) was further transabove.The initial and final ratios of the two competitors were ferred into the cured clones (b 1 p 0 ).Finally constructs were assayed on DM plates with and without phage T5.The relative made where the evolved plasmids were transferred into the fitness (see method of fitness estimation below) of the T5-resisancestral plasmid-free host (b 0 p 1 ).(B) One population of the tant common competitor was 0.992 Ϯ 0.014 relative to the J53-1 plasmid-free E. coli host J53-1 (b 0 ) was propagated by daily ancestor.At least 10 independent competitions were conserial transfer.A random clone from this population was isoducted for each strain.An increased sensitivity can also be lated after 166 days (b 1 *).In addition, constructions were obtained by measuring fitness over several cycles but this made where either of the ancestral plasmids R1 or RP4 was method was avoided to minimize potential effects of plasmid transferred into this evolved host (b 1 *p 0 ).transfer between the competitors.Calculation of Malthusian parameter: Relative fitness (W ) was determined as described by Lenski (1988) and is the ratio of the number of doublings for the tested clone and the resistant filter tips were used in all work associated with transfers common competitor of the evolving populations.The transfers continued for 166 days corresponding to ‫0011ف‬ generations.About every 50 generations the presence of the plasmids was determined on the , appropriate antibiotic plates for 100 colonies from each population.Plasmid isolation was used to verify the presence of plasmids where N i (0) and N j (0) are the initial densities of the tested if resistance markers were absent.One random clone from each clone and the common competitor, respectively, and N i (1) end-population was isolated for further analysis [CD111-116 and N j (1) are their corresponding final densities.(b 1 p 1 ) and CD117 (b 1 *) (Table 1

)]. Clones with new resistance
Determination of relative copy number changes: Minimal phenotypes that had evolved during the experiment were also inhibitory concentrations of ampicillin were determined for isolated (CD139 and CD142-143) and further analyzed when the ancestral and evolved strains as an indirect measure of they represented Ͼ5% of the total population.All clones isolated plasmid copy number (Uhlin and Nordstro ¨m 1977).Drops at the end of the experiment are referred to as evolved in the (10 l) corresponding to 10 2 cells from overnight cultures following text.
were spotted onto plates containing different concentrations Curing and construction of new plasmid-host combinations of ampicillin.The lowest concentration of ampicillin where from evolved clones: The plasmids from the evolved clones were inhibition of growth was detected was determined after overtransferred into the ancestral host J53-1.Transfer was done either night incubation at 37Њ.In a second approach to determine by conjugation in two steps, where E. coli LE392-1 served as an copy number change, DNA band intensities on agarose gels intermediate host, or by transformation using the CaCl 2 procefrom plasmid isolations were measured and expressed as a dure (Maniatis et al. 1989), and resulted in clones CD130-135 ratio to an internal standard in the plasmid isolation.Equal (b 0 p 1 ).The plasmids R1 and RP4 both carry multiple stability volumes from overnight cultures of CD104 (b 0 p 0 ) and CD111systems, including postsegregational killing and partitioning, 113 (b 1 p 1 ) were each mixed with an equal volume from one which makes natural segregant formation very rare.Clones cured single overnight culture of a tester plasmid containing strain of their plasmids (b 1 ) were instead obtained by using two slightly (the relevant characteristic of the tester plasmid is that it different protocols.Plasmid pKG339 ( Jensen et al. 1995), which differs in size from the plasmids under study).Plasmid DNA contains the R1 copA gene under control of a LacI-regulated was isolated from the mixtures and run on an agarose gel.The promoter, was transformed into CD111-113.Transformants were isolated and selected clones were plated on isopropyl thioga-tester plasmid was included to make the assay independent of the DNA extraction efficiency and functioned as a standard Where the ancestral plasmids were transferred to the for quantification.A similar procedure was done for CD105 evolved hosts that were cured of their evolved plasmids (b 0 p 0 ) and CD114-116 (b 1 p 1 ).The gel was then stained for 2 (b 1 p 0 ) we found no significant costs of the plasmids.For hr in ethidium bromide.The intensity of the ancestral or the host that had evolved in the absence of plasmids evolved R1 or RP4 plasmid band was determined and normalized against the intensity of the coextracted tester plasmid reduced costs of the ancestral plasmids (b 1 *p 0 ) were band using Scion Image Version 3b (Scion Corporation, Freddemonstrated, as compared to the costs on the ancestral erick, MD).The relative copy number of each evolved plasmid host (b 0 p 0 ).For the ancestral plasmid R1 on this host is expressed as the ratio of the evolved plasmid and the tester (CD148, b 1 *p 0 ), we found no significant cost and the plasmid band intensities from one coextraction, divided by ancestral plasmid RP4 (CD148, b 1 *p 0 ) had reduced its the ratio of the ancestral plasmid and its corresponding tester plasmid band intensity.Any differences in yield between the cost to 9% (P Ͻ 0.002; Table 2).supplemented with kanamycin and phage T5.
Both plasmids were still present, as determined by plasmid isolation, after several successive platings on media selective for pFF1ts97.

Phenotypic evolution:
Clones that had lost one or Plasmid stability and general fitness changes: Plasmidmore antibiotic resistances emerged in all R1 populafree cells were never detected in the populations tions but represented a minority of the cells (Table 3).founded by R1-or RP4-bearing clones over the ‫0011ف‬ The plasmids from representatives of these minority generations of selection in an antibiotic-free environphenotypes were further analyzed by restriction endoment.The plasmids initially decreased the fitness of E.
nuclease digestion and showed loss of restriction fragcoli J53-1, with 6% (t-test, P Ͻ 0.002) for R1 (CD104, ments of the size corresponding to the respective antibib 0 p 0 ) and 21% (P Ͻ 0.001) for RP4 (CD105, b 0 p 0 ; Table otic-resistance genes (Figure 2A; Clerget et al. 1981).2).The random clones isolated at the end of the experi-Consequently all resistance gene fragments were absent ment (b 1 p 1 ) had increased their fitness, with up to 70% in the digestions of the clones CD139 and CD142 that as compared to their plasmid-bearing ancestors (b 0 p 0 ; had lost all the plasmid resistance phenotypes.Clone Table 2).
CD143 had retained only those resistance gene frag-Compensatory evolution: The respective contribuments containing the kanamycin resistance and its flanktions of host-and plasmid-encoded mutations to the ing insertion sequence (IS1) elements.general fitness increase observed in the evolved clones Several changes in the evolved plasmids from the (b 1 p 1 ) were measured on combinations of ancestral and randomly drawn clones were also revealed.The RP4 evolved hosts and plasmids.
isolates CD114 and CD115 (b 1 p 1 ) differed from the an-The costs of the evolved plasmids were tested in both cestral plasmid in their restriction endonuclease digesthe evolved and ancestral host backgrounds.With one tion patterns (Figure 2B) and showed several novel reexception (see below, clone CD122) there were no sigstriction sites in the transfer and replication regions nificant differences in fitness between the evolved (data not shown).These changes were probably associclones (b 1 p 1 ) and the evolved clones cured of their plasated with transposition of chromosomally located IS mids (b 1 ).The cost of the evolved plasmids on the anceselements into the plasmids.For the plasmids in both tral host (b 0 p 1 ) was further tested and showed that five CD114 and CD115 possible insertions into the trbE reof the six evolved plasmids had a lower cost than the gion that is involved in pilus production were found corresponding ancestral plasmid (b 0 p 0 ).In contrast, one (Pansegrau et al. 1994).An additional insertion in the of the evolved plasmids, the R1 derivative in clone oriV region of the plasmid in CD115 was also found.CD112, showed an increased cost on the ancestral host One structural change was detected in the random R1 (CD131, 17%, P Ͻ 0.001).
clones as CD113 showed a reduction in size of the restric-The effect of host evolution was assessed by measuring tion fragment that in the ancestral R1 plasmid harbors one IS1 copy and a kanamycin resistance gene (Figure the cost of the ancestral plasmids on the evolved hosts.seen in the ancestral host (b 0 p 1 ; Table 4).Transfer of the ancestral plasmid R1 was also measured from the host that had evolved in the absence of plasmids, CD148 2A).The clone CD113 (b 1 p 1 ) was kanamycin resistant (b 1 *p 0 ).This transfer rate was of the same magnitude as and the size reduction was probably due to excision of transfer from the ancestral host (b 0 p 0 ; Table 4).the IS1 element.
The values obtained by the two different methods The evolved RP4 plasmids in clones CD114 and used for copy number measurements did not fully corre-CD115 were transfer deficient in both the evolved and late.Only when both methods showed a similar deviathe ancestral host.The evolved RP4 plasmid in clone tion in copy number was an average of these values used CD116 had retained its transferability but the transfer rates of this plasmid and of the ancestral RP4 could not to assign changes in copy number.The RP4 derivative be determined for the liquid culture conditions used in clone CD116 (b 1 p 1 ) showed an increase in copy numin these experiments, as they did not exceed the transfer ber of ‫%07ف‬ and the R1 clone CD111 (b 1 p 1 ) showed a that occurred after plating on the selective media.The 20% reduction in copy number (Table 5).evolved R1 clones (b 1 p 1 ) showed lower transfer rates compared to the ancestral R1 clone (b 0 p 0 ).The reduc-DISCUSSION Many antibiotic-resistance genes are carried by conjugative plasmids but studies to date on compensatory evolution of antibiotic resistance have focused on genes encoded either on chromosomes or on nonconjugative plasmids (Bouma and Lenski 1988;Modi and Adams 1991;Shrag and Perrot 1996;Bjo ¨rkman et al. 1998Bjo ¨rkman et al. , 2000;;Reynolds 2000;Nagaev et al. 2001).We examined how the cost of carrying conjugative antibioticresistance plasmids is affected by evolution in an antibiotic-free environment.Parallel populations of an E. coli host carrying the natural conjugative plasmids R1 or RP4 were propagated for ‫0011ف‬ generations by serial transfer.The cost associated with plasmid carriage before and after experimental evolution was determined in competition experiments using a plasmid-free common competitor.Both plasmids R1 and RP4 imposed fitness costs on E. coli J53-1 at the start of the experiment (Table 2).At the end of the experiment all tested clones showed an increased fitness compared to the respective pected to account for part of this fitness increase but  If compensatory mutations on the evolved plasmids were alone responsible for the decreased cost one would a MIC, minimal inhibitory concentration in micrograms per expect that the ancestral plasmids would still impose a milliliter of ampicillin, based on three independent measurements.
similar cost on the evolved hosts as on the ancestral b Relative change in plasmid DNA content compared to host.This was, however, not seen; instead we found that respective ancestral plasmid as determined by agarose gel electhe ancestral plasmids had no significant cost on the trophoresis and density measurements of relative band intensievolved hosts that were cured of their evolved plasmids ties.Average of two independent experiments is shown.
(b 1 p 0 ; Table 2).This result indicates that additional compensatory mutations had occurred in the chromosome.The surprising aspect of these results lies in the apparent changes in response to the cost of plasmid carriage were also anticipated.
redundancy of the compensatory mutations shown by the compensatory mutations occurring in both the plas-Reduction of the cost of plasmid carriage can be accomplished through two main processes: segregation mid and the host.When a mutation compensatory for the cost of plasmid carriage occurs on either the host and compensatory mutation.For plasmid-borne antibiotic resistance segregation will lead to the loss of resis-or the plasmid replicon, selection for a second compensatory mutation on the second replicon should be abol-tance by the formation of plasmid-free cells whereas compensatory mutations may or may not abolish resis-ished.
The host that evolved in the absence of plasmids (b 1 *) tance.Both segregation and compensatory mutation are likely to occur in a large population, resulting in a race expectedly demonstrated an increase in fitness in response to the growth conditions.However, a second between the two.Under conditions of antibiotic-resistance selection, compensatory mutation is primarily surprising result showed that this well-adapted host naive to plasmid carriage was more tolerant to the cost of available, but under conditions where antibiotic selection is released it is unclear which of these processes a plasmid.For the ancestral plasmid R1 on this host we found no significant cost and the ancestral plasmid RP4 will dominate.
Like most natural conjugative plasmids, the plasmids had greatly reduced its cost (Table 2).As an explanation for the results on compensatory used in this study have stability systems such as postsegregational killing and partitioning that make their segre-evolution we suggest that during the experiment both the host and the plasmids acquire mutations that are gation rates very low (Nordstro ¨m et al. 1980;Schmidhauser and Helinski 1985).The fact that plasmid-free selected but for different purposes.The plasmids evolve compensatory mutations that are selected because they cells were never detected in the evolved populations was most likely affected by the action of these stability enhance fitness of the host whereas the host evolves toward the general growth conditions with a side effect systems.Earlier studies have also shown that plasmid RP4 can be stably maintained in continuous culture of decreasing the cost of plasmid carriage.If the plasmid evolves compensatory mutations more rapidly than the (Melling et al. 1977;Jones et al. 1980).There is a possibility that segregants were reinfected if they ap-host evolves mutations with compensatory side effects this would result in the apparent redundancy of com-peared, but the very low transfer rates we found under these conditions (see below) makes this explanation pensatory mutations.It would also explain why a host that has evolved in the absence of plasmids suffers a unlikely.
Since segregation was not detected in these experi-lower cost of an ancestral plasmid than an ancestral host does.This explanation is also consistent with the results ments we directed our efforts to examine the effect of compensatory evolution.The first indication of com-that an evolved plasmid does better than its ancestor in the ancestral host.pensatory mutations was seen when the fitness of plasmid-free clones (b 1 ) constructed from evolved plasmid- The cured clone CD122 showed a higher fitness when its plasmid had been reintroduced compared to the bearing clones (b 1 p 1 ) was measured.These measurements showed no difference in fitness between the pairs of fitness of the plasmid-bearing clone before curing in the control experiment.This suggests that chromosomal plasmid-carrying (b 1 p 1 ) and plasmid-free (b 1 ) clones with genetic changes occurred during the curing procedure Antibiotic-resistance expression is a third type of plasmid function that can be altered to reduce cost.Most and the fitness values measured for the cured state of this clone are uncertain.A reasonable explanation for of the clones from the evolved populations had retained all their antibiotic resistances and had obviously re-this result is that a mutualistic association resulting in strong plasmid dependency had evolved in this clone.
duced their costs by other means.However, R1-carrying populations did contain minor fractions of clones that We believe that the strong selection applied when the resident plasmid was forced out by an incompatible had lost resistances.The antibiotic sensitivities of these evolved clones correlated with deletions of the resis-replicon during the curing procedure selected for a mutation compensatory for the loss of the plasmid.
tance genes and included loss of all the plasmid-borne resistances.Similar results have previously been pre- Bouma and Lenski (1988) and Modi and Adams (1991) previously showed that evolution leading to plasmid de-sented for the conjugative plasmid TP120, where loss of resistances was observed after selection in chemostats pendency occurred during experimental evolution of small nonconjugative plasmids.They were able to dem-(Godwin and Slater 1979).
Conclusions: These results demonstrate that the fit-onstrate, by the isolation of plasmid-free cells by natural segregation that had a significantly lower fitness com-ness cost of conjugative resistance plasmid carriage can be reduced and is often completely ameliorated after pared to the plasmid-bearing cells, that plasmid dependency had evolved.
selection in an antibiotic-free environment.Reversion to sensitivity by plasmid segregation was not observed What was the basis for the decreased costs of the evolved plasmids from five out of six independent in these experiments.Instead, compensatory evolution was the major pathway.As previously suggested a de-clones?First one may consider conjugation.Plasmid RP4 has a constitutively expressed transfer function creased usage of antibiotics may hence not be sufficient for dealing with the antibiotic-resistance problem as whereas R1 is repressed for transfer.One part of the cost may be associated with pilus production.We found sensitive bacteria are prevented from invading resistant bacterial populations after compensatory evolution has that two of the evolved RP4 isolates were transfer deficient and that this deficiency correlated with changes occurred.However, the emergence of plasmids with deletion of all antibiotic-resistance markers from the in their restriction endonuclease digestion patterns in genes involved in pilus production.The evolved R1 basic replicon in some of our experiments shows that there is still a potential for selection of antibiotic-sensi-clones (b 1 p 1 ) showed lower transfer rates than the ancestral R1 clone (b 0 p 0 ).This reduction in transfer rate was tive populations.A consequence of the plasmidencoded compensatory mutations is that the resistance likely due to host evolution resulting in a host-directed suppression of transfer since the reduced transfer of genes associated with these plasmids may have an enhanced potential to invade new populations by hori-these plasmids was observed only in the evolved host (b 1 p 1 ) and not in the ancestral host (b 0 p 1 ; Table 4).The zontal transfer as compared to the more limited chromosomal compensatory mutations that still are confined transfer rate of the ancestral R1 from the host that had evolved in the absence of plasmids (b 1 *p 0 ) was of the to one population.It remains to be elucidated if plasmid compensatory mutations are specific for the strain they same magnitude as transfer from the ancestral host (b 0 p 0 ).This result supports the idea that the mechanism were selected in or if they also reduce the cost of the plasmid carriage when it transfers to other strains or responsible for the reduced transfer rates in the former case evolved in direct response to plasmid carriage (Ta-species.ble 4).We did not attempt to localize the chromosomal cost reduction (Modi and Adams 1991).The copy number has been shown to be ‫/5-4ف‬cell for plasmid R1 and 4-6 for plasmid RP4 (Figurski et al. 1979; Nord- we could observe a 20% reduction in copy number 7: 1513-1523. (Table 5).In contrast, an increased copy number corre- compatible with the original replicon.
ancestral and evolved plasmids in the overnight cultures were Control experiments where the evolved plasmid was first corrected for.reintroduced into the cured host (i.e., reconstruction Phenotypic and molecular analysis of evolved plasmids: The of b 1 p 1 from b 1 ) showed that the curing procedure did resistance phenotype of 100 isolates from each end population not affect fitness (data not shown) with one exception.was determined by replica plating on antibiotic-containing media.Restriction enzyme analysis was performed on plasmid This cured clone CD122 showed a higher fitness when DNA that was isolated by alkaline lysis (Birnbom and Doly its plasmid had been reintroduced compared to the 1979) and digested with either EcoRI (R1) or EcoRV and Pst I fitness of the plasmid-bearing clone before curing.(RP4).Digestions were performed according to the manufac-A second problem we encountered when curing the turer's recommendations (New England Biolabs, Beverly, MA) evolved clones was observed in the evolved RP4-bearing and separated on 0.9% agarose TAE gels.The rate of conjugal transfer(Simonsen et al. 1990) was determined for evolved clone CD116 (b 1 p 1 ).During curing when this clone was and ancestral plasmids in evolved and ancestral hosts in experitransformed with the plasmid pFF1ts97 that carries the ments equal to one 24-hr cycle of serial transfer.The number same basic replicon as the ancestral RP4, the expected of transconjugant CD100 cells was determined on DM plates loss of the resident evolved RP4 plasmid was not seen.
were measured as the fitness relative to the common competitor CD100.Ϯ standard error.All fitness values were corrected for the effect of the T5-resistance marker in CD100.a Standard error, number of experiments.b Fitness measurement is the same as for population 1. c See results for details.d This fitness measurement is the same as for population 4. e ND, not determined.

Figure 2 .
Figure 2.-Restriction fragments of ancestral and evolved

CD104
these results.The costs of five of the six tested evolved plasmids on the ancestral host Strain MIC a Relative DNA content b (b 0 p 1 ) were lower than the costs of the corresponding We thank Dan Andersson, Diarmaid Hughes, Bruce Levin, Sophie mutations responsible for this effect but several host-Maisnier-Patin, and two anonymous reviewers for helpful comments and discussions; Kenn Gerdes, Donald Helinski, Daniel Dykhuizen, encoded factors that influence plasmid transfer have and Michael Travisano for the gifts of plasmids and phage; and Kamal been identified in previous studies.For example, the Gandhi for excellent technical assistance.This work was funded by ArcA protein, involved in cellular redox sensing, affects grants from the Swedish Research Council and The Royal Swedish R1 transfer (Strohmaier et al. 1998).Academy of Sciences to C.D and from the National Institutes of Gen-Copy number reduction is another possible route for eral Medical Sciences and National Institutes of Health to L.C.
Bjo ¨rkman, J., D. Hughes and D. I. Andersson, 1998 Virulence of sponding to 7-10 copies per cell was observed for the antibiotic-resistant Salmonella typhimurium.Proc.Natl.Acad.Sci.USA 95: 3949-3953.evolved RP4 clone CD116 (Table 5).The mutation that Bjo ¨rkman, J., I. Nagaev, O. G. Berg, D. Hughes and D. I. Andersson, made this plasmid increase its copy number could ac-2000 Effects of environment on compensatory mutations to count for our failure to cure this clone by making it ameliorate the cost of antibiotic resistance.Science 287: 1479-1482.

TABLE 1 Escherichia coli strains and plasmids
R  Jensen et al. (1995)a In strain designations b refers to the host and p to the plasmid (see Figure1legend for details).b Nal, nalidixic acid; Cm, chloramphenicol; Km, kanamycin; Ap, ampicillin; Sm, streptomycin; Tc, tetracycline.c NCTC, National Collection of Type Cultures, London.

TABLE 5
one exception (see below, clone CD122; Table2).Compensatory mutations in either the chromosome or the