BRIEF COMMUNICATION

doi:10.1111/j.1558-5646.2011.01531.x

TRAGEDY OF THE COMMONS AMONGANTIBIOTIC RESISTANCE PLASMIDSJeff Smith1,2,3

1Department of Biology, Emory University, Atlanta, Georgia 303022E-mail: jeffsmith@wustl.edu.

Received July 28, 2011

Accepted November 21, 2011

As social interactions are increasingly recognized as important determinants of microbial fitness, sociobiology is being enlisted

to better understand the evolution of clinically relevant microbes and, potentially, to influence their evolution to aid human

health. Of special interest are situations in which there exists a “tragedy of the commons,” where natural selection leads to a

net reduction in fitness for all members of a population. Here, I demonstrate the existence of a tragedy of the commons among

antibiotic resistance plasmids of bacteria. In serial transfer culture, plasmids evolved a greater ability to superinfect already-

infected bacteria, increasing plasmid fitness when evolved genotypes were rare. Evolved plasmids, however, fell victim to their

own success, reducing the density of their bacterial hosts when they became common and suffering reduced fitness through

vertical transmission. Social interactions can thus be an important determinant of evolution for the molecular endosymbionts of

bacteria. These results also identify an avenue of evolution that reduces proliferation of both antibiotic resistance genes and their

bacterial hosts.

KEY WORDS: Cheating, conflict, cooperation, horizontal gene transfer, selfish genetic elements, symbiosis, virulence.

The tragedy of the commons is an economic analogy that describes

how individuals, acting rationally and in their own self-interest,

can overexploit a shared resource even when it is in no one’s

long-term interest to do so (Hardin 1968). It was introduced to

contrast with Adam Smith’s claim that economic self-interest

leads individuals, as if “led by an invisible hand,” to promote

the public interest (Smith 1776). The analogy has been imported

into evolutionary biology, where it refers to a situation in which

competition reduces the resource over which individuals compete,

leading to a reduction in fitness for all members of the population

or group (reviewed in Dionisio and Gordo 2006; Rankin et al.

2007). The net effect is that, somewhat counterintuitively, natural

selection reduces population-mean fitness. Rankin et al. (2007)

distinguish between “collapsing” tragedies in which a resource

is entirely depleted and “component” tragedies that reduce mean

3Current address: Department of Biology, Washington University in

St. Louis, Campus Box 1137, One Brookings Drive, St. Louis, Missouri

63130–4899.

group fitness but fall short of total collapse. Because avoiding a

tragedy seems to require that individuals give up personal fitness

to benefit the larger group, the analogy is often invoked in the

context of the evolution of cooperation.

The potential for tragedies of the commons among

pathogenic microbes has been appreciated for some time. The-

oretical models have proposed that competition among pathogen

genotypes within a host allows aggressively reproducing strains

to gain short-term advantage but may also shorten the duration of

infection due to increased host death or immune clearance. The

host is thus a resource that is shared among coinfecting strains,

and optimal infectious transmission may require cooperative re-

straint (Lewontin 1970; Axelrod and Hamilton 1981; Levin and

Pimental 1981; Bremermann and Pickering 1983; Nowak and

May 1994; May and Nowak 1995; Frank 1996). Another kind of

social dilemma occurs when the success of microbial infections

depends on the production of costly diffusible products whose

benefits can also be obtained by nonproducing “cheater” strains

(Brown 1999; Chao et al. 2000; Smith 2001; West et al. 2007;

1 2 6 9C© 2012 The Author(s). Evolution C© 2012 The Society for the Study of Evolution.Evolution 66-4: 1269–1274

BRIEF COMMUNICATION

Buckling and Brockhurst 2008). Microbial sociobiology may

thus be able to help us better understand how selection acts on

pathogenic traits and to predict whether pathogen evolution in

response to medical interventions will lead to desirable or unde-

sirable outcomes (Foster 2005; Kohler et al. 2010). Some have

even suggested that we may be able to influence pathogen evo-

lution to aid human health, for example, by using intervention

strategies that select for less-virulent pathogens (Dieckmann et

al. 2002) or introducing cheater strains to sabotage infections

(Brown et al. 2009; Marriott and Dimmock 2010).

Here, I demonstrate the existence of a tragedy of the com-

mons in a clinically relevant system at an even smaller level of bi-

ological organization—among the plasmids of bacteria. Plasmids

are mobile genetic elements that often carry genes for resistance

to antibiotics and are major determinants of resistance among bac-

terial pathogens and opportunistic infections (Frost et al. 2005; Su

et al. 2008). Plasmids reproduce independently of chromosomes

and are typically inherited by both daughter cells after division.

Many plasmids can also transmit themselves infectiously among

cells. Plasmid fitness thus has two components: one through ver-

tical transmission and another through horizontal transmission.

Because plasmid ecology and evolution is somewhat decoupled

from that of their bacterial hosts (Boyd and Hartl 1997), with both

positive and negative effects on host fitness (Bouma and Lenski

1988; Modi and Adams 1991; Turner et al. 1998; Dahlberg and

Chao 2003; Dionisio et al. 2005), plasmids can be thought of as

molecular endosymbionts of bacteria.

I previously described an experiment in which plasmids

evolved an increased ability to “superinfect” bacteria already in-

fected with incompatible plasmids (Fig. 1A; smith 2011). Here,

I test theoretical predictions that within-host competition caused

by superinfection creates a tragedy of the commons. I measure the

fitness of evolved plasmids in direct competition with the ances-

tral genotype, including both vertical and horizontal transmission

components of fitness, and measure how evolved plasmids affect

host abundance.

Materials and MethodsCULTURE CONDITIONS

All bacteria were cultured in 10 mL Davis minimal media sup-

plemented with 1000 μg/mL glucose and 20 μg/mL uracil, in-

cubated at 37◦C in 50 mL conical flasks shaking at 200 rpm.

Where indicated, cultures were passaged with 100-fold dilution

(0.1 mL into 9.9 mL fresh media). These are the same conditions

as the previous evolution experiment (smith 2011). Cell densi-

ties were estimated from colony counts on tetrazolium arabinose

(TA) plates (Levin et al. 1977), on which Ara+ colonies appear

white and Ara− colonies appear red. The evolution experiment

and fitness assays were performed in antibiotic-free media, but

Figure 1. (A) Plasmids evolved in serial transfer culture (Smith

2011). All host bacteria were infected, but plasmids can spread

by superinfecting already-infected hosts (inset). (B) An evolved

plasmid genotype invades populations of the ancestral plasmid

through superinfection. The evolved plasmid is highly virulent,

slowing growth of infected hosts and reducing population density

when it becomes common. Solid line: cells infected with ancestral

plasmid. Dashed line: introduced donor strain of hosts infected

with evolved plasmid. Dotted line: transconjugant cells that have

acquired evolved plasmids via superinfection. Lines and error bars

are geometric mean ± range of data (n = 4). (C) In control com-

petitions between differently marked versions of the ancestral

plasmid, some superinfection occurs but transconjugants remain

rare.

antibiotic resistances were used as genetic markers of plasmid

and host genotype. Where indicated, antibiotics were included

in plates at the following concentrations: kanamycin (Km) at

25 μg/mL, chloramphenicol (Cm) at 25 μg/mL, rifampin (Rif) at

10 μg/mL, and nalidixic acid (Nal) at 2.0 μg/mL.

BACTERIA STRAINS AND PLASMIDS

All bacterial strains used here are derivatives of Escherichia coli

strain MG1655 obtained from the National Type Culture Collec-

tion. NalR and RifR strains were obtained as spontaneous mutants

after plating on Nal or Rif TA medium, respectively. Plasmid

R1–19 (Meynell and Datta 1967) was obtained from B. Levin

1 2 7 0 EVOLUTION APRIL 2012

BRIEF COMMUNICATION

(Emory University). Plasmid R1–19K, a CmS mutant of R1–19,

was obtained from K. Nordstrom (Uppsala University).

Evolved plasmids come from a previously described exper-

iment in which hosts were infected with R1–19 and passaged in

daily serial transfer culture (smith 2011). In that paper’s termi-

nology, evolved plasmids were sampled from population N1 after

100 generations of evolution. This population was chosen because

it contains evolved plasmids that retained resistance to chloram-

phenicol. The original evolution experiment allowed both plas-

mids bacteria to evolve. The plasmids studied here were among

the very first evolved genotypes to appear. They appeared very

rapidly in the original experiment (∼14 days), before evolved host

genotypes were detected or would be expected to reach apprecia-

ble frequency. These plasmids, then, evolved primarily in the

ancestral host genotype. Plasmids from later on in the evolution

experiment would be more likely to have evolved in evolved hosts.

Individual plasmid genotypes were sampled from the evolved

population and moved onto a common ancestral host background.

To accomplish this, frozen stocks of the evolved population (Ara−)

and an ancestral plasmid-free host (js171; Ara+ RifR) were sep-

arately grown for one day then coinoculated into fresh media

and grown another day, during which evolved plasmids infected

plasmid-free hosts. These cultures were then plated onto TA Rif

Cm to select for ancestral hosts infected with an evolved plasmid.

Random colonies were resuspended in saline, used to inoculate

each of four flasks, and incubated for one day. These four cul-

tures were then used as replicate test strains in the fitness assays

described below. As a control, test strains were also made using

ancestral plasmids from an ancestral host strain (js42; Ara− NalR)

instead of the evolved population.

PLASMID ASSAYS

Plasmid fitness was measured by competing plasmids in serial

passage culture against a differently marked version of the an-

cestral plasmid. For these assays, test strain cultures were diluted

1000-fold in saline, mixed at equal volume with overnight cultures

of an ancestral host infected with the competitor plasmid (js244;

Ara− NalR R1–19K), and passaged one day. Some genotypes

were passaged for additional days to follow plasmid/host popu-

lation dynamics. Total cell densities were calculated from colony

counts on TA plates. Densities of all infected cells were calculated

from colony counts on TA Km plates. Densities of donor strains

for evolved plasmids were calculated from Ara+ colony counts

on TA Cm or TA Rif Cm plates. Densities of transconjugant cells

newly infected with the evolved plasmid were calculated from

Ara− colony counts on TA Cm or TA Nal Cm plates. Densities

of cells infected with the differently marked ancestral plasmid

were calculated as total infected density minus transconjugant

density.

A mathematical model of population dynamics predicted in

the fitness assays is included as Supporting Information. Plas-

mid fitness was calculated in terms of absolute fitness per serial

transfer. If n is the density of cells infected with a given genotype

immediately after dilution and n′ is the density of cells infected

with that same genotype after one day’s growth, then the ab-

solute fitness of that genotype over the whole growth cycle is

w = n′/n. The vertical transmission component of plasmid fitness

was measured by calculating w using only the densities of the

donor strain for tested plasmids (i.e., excluding transconjugants).

Relative fitness was calculated as the ratio of tested to competitor

fitness wtest/wcompetitor.

To measure how plasmids affect host populations, test strains

were passaged one day. Population densities were calculated from

colony counts on TA plates.

As an additional measure of plasmid fitness through vertical

transmission, test strains were competed against plasmid-resistant

strain js177 (smith 2011). Against a resistant host, plasmid fitness

is primarily determined by its effect on the survival and reproduc-

tion of hosts who have inherited their infections vertically. For

these assays, overnight cultures of test strains and resistant hosts

were mixed at equal volume and passaged one day. Densities

of plasmid-infected test hosts were calculated from Ara+ colony

counts on TA or TA Rif plates. Densities of resistant hosts were

calculated from Ara− colony counts on TA plates. Fitness were

calculated as above, only now with the resistant host strain as the

competitor.

STATISTICS

All data were analyzed using R 2.8.1 (R Development Core Team

2011). Statistical models were fit to log10 relative fitness or log10

host density. Differences between differently marked versions of

the ancestral plasmid were tested using two-tailed one-sample

t-tests (t.test procedure). To test differences between the ancestral

and evolved plasmids as a group, data were fit to linear mixed

effects models by maximum likelihood (lme procedure). Signif-

icance was determined using likelihood ratio tests on models fit

with or without ancestral/evolved status as a fixed effect. Plasmid

genotype was modeled as a random effect nested within ances-

tral/evolved status, allowing evolved genotypes to vary in fitness

in addition to an overall difference between evolved plasmids

and the ancestor. Differences between individual evolved plas-

mids and the ancestor were tested using two-sample t-tests (t.test

procedure). These tests were one-tailed because of a prior expec-

tation for the direction of effects under a tragedy of the commons.

To correct for multiple comparisons, tests were only considered

significant below the Bonferoni cutoff level for four comparisons

of P = (0.05/4) = 0.0125.

EVOLUTION APRIL 2012 1 2 7 1

BRIEF COMMUNICATION

ResultsWhen I introduced evolved plasmids at low frequency into popula-

tions of bacterial hosts infected with a differently marked version

of the ancestral plasmid, evolved plasmids rapidly increased in

frequency by superinfecting hosts previously infected with the

ancestral plasmid genotype (Fig. 1B shows one example). In con-

trol competitions between two differently marked ancestral plas-

mid genotypes, very little superinfection occurred (Fig. 1C). I

calculated the relative fitness of the ancestral and four evolved

plasmid genotypes over the first transfer of these experiments,

when evolved plasmids were still rare (Fig. 2A). In the control

competitions, the ancestral plasmid was slightly but significantly

less fit than its competitor (t = −9.8, df = 3, P = 0.0022), possi-

bly due to the different genetic markers. Evolved plasmids were

much more fit than the ancestor when compared together as a

group (LR = 9.2, P = 0.0024) or individually (E1: t = 11.4, df =3.0, P = 6.8 × 10−4; E2: t = 16.3, df = 3.1, P = 2.0 × 10−4; E3:

t = 50.7, df = 3.7, P = 1.1 × 10−6; E4: t = 12.5, df = 3.1, P =4.5 × 10−4).

Over the course of several passages, evolved plasmids in-

creased in frequency. When they became common, cultures

were visibly delayed becoming turbid (an indicator of bacterial

density), and host population densities were severely reduced

(Fig. 1B). To investigate this effect further, I separately cultured

hosts infected with either evolved or ancestral plasmid genotypes

(Fig. 2B). Compared to the ancestral plasmid, evolved plasmids as

a group reduced host densities by one or two orders of magnitude

(LR = 7.7, P = 0.0057). Compared individually, two evolved

genotypes showed significant reduction (E2: t = −10.2, df = 3.1,

P = 8.3 × 10−4; E4: t = −4.6, df = 3.0, P = 0.0094) whereas

the other two were only significant before correcting for multiple

comparisons (E1: t = −3.1, df = 3.0, P = 0.025; E3: t = −3.0,

df = 3.0, P = 0.028).

Although the overall frequency of evolved plasmids

increased in competition experiments, the frequency of the orig-

inally infected host strain decreased (Fig. 1B). In ancestral con-

trols, the frequency of these “donor” hosts changed relatively lit-

tle (Fig. 1C). I estimated the vertical transmission component of

plasmid fitness from the change in donor frequency over the first

transfer (Fig. 2C). Compared as a group to the ancestral control,

evolved plasmids had reduced fitness through vertical transmis-

sion (LR = 4.2, P = 0.039). Compared individually, two had

significantly reduced fitness (E2: t = −8.3, df = 3.2, P = 0.0015;

E4: t = −4.5, df = 3.1, P = 0.0096), one was only significant

before correcting for multiple comparisons (E3: t = −2.5, df =3.2, P = 0.041), and one was almost significant (E1: t = −2.2,

df = 3.0, P = 0.058).

As another measure of plasmid fitness through vertical trans-

mission, I competed infected hosts against a bacterial strain

Figure 2. Plasmid evolution creates a tragedy of the commons.

(A) Evolved plasmids are more fit than their ancestor when rare

but (B) reduce host population density when common. (C) Evolved

plasmids have reduced fitness through vertical transmission when

measured against the ancestral plasmid genotype or (D) a plasmid-

resistant host strain. Points: replicate fitness measurements. Bars:

geometric mean of replicates for each plasmid genotype (n = 4).

Evolved plasmid E4 is the genotype shown in Figure 1B.

resistant to plasmid infection (Fig. 2D). In these experiments, host

strains infected with evolved plasmids were less fit than those

infected with the ancestral plasmid when compared as a group

(LR = 6.1, P = 0.014) or when compared individually (E1: t =−4.0, df = 3.5, P = 0.011; E2: t = −8.2, df = 5.7, P = 1.1 ×10−4; E3: t = −7.2, df = 5.1, P = 3.8 × 10−4; E4: t = −5.6,

df = 4.4, P = 0.0018).

1 2 7 2 EVOLUTION APRIL 2012

BRIEF COMMUNICATION

DiscussionA tragedy of the commons occurs when natural selection favors

traits that diminish the resource over which individuals compete,

leading to a net reduction in fitness for all individuals (Rankin

et al. 2007). By this standard, plasmid evolution in these ex-

periments followed a tragedy of the commons. Greater ability

to superinfect already-infected bacteria increased the fitness of

evolved plasmids when rare and directly competing with the an-

cestral type. Evolved plasmids, however, fell victim to their own

success. When they were common, they severely reduced the den-

sity of the bacterial hosts on which they depend. They also had

reduced fitness through vertical transmission, the majority com-

ponent of fitness when plasmid genotypes are locally common.

Selection favored plasmid traits, therefore, that led to a net de-

crease in population-mean fitness. Tragedies of the commons can

therefore occur among the mobile genetic elements of bacteria

that so often carry and disseminate genes for antibiotic resistance

and pathogen virulence (Frost et al. 2005).

In this system, the tragedy follows from plasmids failing to

use hosts prudently. Abolishing plasmid-borne resistance to su-

perinfection can cause membrane damage and host death when

plasmids are common (Ou 1980). Following the classification

system of Rankin et al. (2007), hosts are a preexisting resource

whose optimal usage requires cooperative restraint. In this sense,

the tragedy is similar to those predicted to exist in host/pathogen

systems (Lewontin 1970; Axelrod and Hamilton 1981; Levin and

Pimental 1981; Bremermann and Pickering 1983; Frank 1996;

Read and Taylor 2001). Evidence for such a tragedy is lack-

ing among human pathogens (Read et al. 2002), but similar ef-

fects have been seen in nonpathogenic microbes (MacLean 2008;

Eshelman et al. 2011).

A different form of social dilemma occurs when individuals

compete by producing less of a costly public good (West et al.

2007; Buckling and Brockhurst 2008). The Pseudomonas bacteria

that infect cystic fibrosis patients, for example, evolve to produce

fewer secreted products that increase proliferation within hosts

(Kohler et al. 2009; Rumbaugh et al. 2009; Kohler et al. 2010).

Viral infection success can also be reduced by “defective inter-

fering particles” that produce less of a gene product necessary

for replication but can use products produced by other viruses

(Marriott and Dimmock 2010). So far, there is no evidence for a

public goods dilemma among the plasmids in these experiments.

If the primary effect of superinfection, for example, was to favor

defective interfering genotypes, one would expect evolved plas-

mids to impose less of a metabolic burden on hosts, increasing

host density and increasing fitness through vertical transmission.

Two clinically desirable outcomes were seen in the evo-

lution experiments from which these plasmids were isolated—

reduced bacterial proliferation and rapid loss of antibiotic resis-

tance (smith 2011). Other studies of plasmid evolution, however,

have often found that resistance persists long after antibiotic se-

lection is removed (Bouma and Lenski 1988; Modi and Adams

1991; Turner et al. 1998; Dahlberg and Chao 2003; Dionisio

et al. 2005). Future work to clarify the genetic and environmental

causes of these different outcomes could help us understand how

to enlist evolution as an aid, rather than a hindrance, in the fight

against antibiotic resistant bacteria.

ACKNOWLEDGMENTSD. Queller, D. Rozen, and J. Strassmann provided helpful discussion andcomments on the manuscript. This research was supported by NIH grantGM33782–17 to B. Levin (Emory University).

LITERATURE CITEDAxelrod, R., and W. D. Hamilton. 1981. The evolution of cooperation. Science

211:1390–1396.Bouma, J. E., and R. E. Lenski. 1988. Evolution of a bacteria/plasmid associ-

ation. Science 335:351–352.Boyd, E., and D. Hartl. 1997. Recent horizontal transmission of plasmids be-

tween natural populations of Escherichia coli and Salmonella enterica.J. Bacteriol. 179:1622–1627.

Bremermann, H. J., and J. Pickering. 1983. A game-theoretical model ofparasite virulence. J. Theor. Biol. 100:411–426.

Brown, S. P. 1999. Cooperation and conflict in host-manipulating parasites.Proc. R. Soc. Lond. B 266:1899–1904.

Brown, S. P., S. A. West, S. P. Diggle, and A. S. Griffin. 2009. Social evolutionin micro-organisms and a Trojan horse approach to medical interventionstrategies. Philos. Trans. R. Soc. Lond. B 364:3157–3168.

Buckling, A., and M. A. Brockhurst. 2008. Kin selection and the evolution ofvirulence. Heredity 100:484–488.

Chao, L., K. A. Hanley, C. L. Burch, C. Dahlberg, and P. E. Turner. 2000. Kinselection and parasite evolution: higher and lower virulence with hardselection and soft selection. Q. Rev. Biol. 75:261–275.

Dahlberg, C., and L. Chao. 2003. Amelioratiion of the cost of conjuga-tive plasmid carriage in Eschericha coli K12. Genetics 165:1641–1649.

Dieckmann, U., J. A. J. Metz, M. W. Sabelis, and K. Sigmund. 2002. Adaptivedynamics of infectious diseases. Cambridge Univ. Press, Cambridge.

Dionisio, F., I. C. Conceicao, A. C. R. Marques, L. Fernandes, and I. Gordo.2005. The evolution of a conjugative plasmid and its ability to increasebacterial fitness. Biol. Lett. 1:250–252.

Dionisio, F., and I. Gordo. 2006. The tragedy of the commons, the public goodsdilemma, and the meaning of rivalry and excludability in evolutionarybiology. Evol. Ecol. Res. 8:321–332.

Eshelman, C. M., R. Vouk, J. L. Stewart, E. Halsne, H. A. Lindsey, S. Schnei-der, M. Gualu, A. M. Dean, and B. Kerr. 2011. Unrestricted migrationfavours virulent pathogens in experimental metapopulations: evolution-ary genetics of a rapacious life history. Philos. Trans. R. Soc. Lond. B365:2503–2513.

Foster, K. R. 2005. Hamiltonian medicine: why the social lives of pathogensmatter. Science 308:1269–1270.

Frank, S. A. 1996. Models of parasite virulence. Q. Rev. Biol. 71:37–78.Frost, L. S., R. Leplae, A. O. Summers, and A. Toussaint. 2005. Mobile genetic

elements: the agents of open source evolution. Nat. Rev. Microbiol.3:722–732.

EVOLUTION APRIL 2012 1 2 7 3

BRIEF COMMUNICATION

Hardin, G. 1968. The tragedy of the commons. Science 162:1243–1248.Kohler, T., A. Buckling, and C. van Delden. 2009. Cooperation and virulence

of clinical Pseudomonas aeruginosa populations. Proc. Natl. Acad. Sci.USA 106:6339–6344.

Kohler, T., G. G. Perron, A. Buckling, and C. van Delden. 2010. Quorumsensing inhibition selects for virulence and cooperation in Pseudomonas

aeruginosa. PLoS Path. 6:e1000883.Levin, S., and D. Pimental. 1981. Selection of intermediate rates of increase

in parasite-host systems. Am. Nat. 117:308–115.Levin, B. R., F. M. Stewart, and L. Chao. 1977. Resource-limited growth, com-

petition, and predation: a model and experimental studies with bacteriaand bacteriophage. Am. Nat. 111:3–24.

Lewontin, R. C. 1970. The units of selection. Annu. Rev. Ecol. Syst. 1:1–18.MacLean, R. C. 2008. The tragedy of the commons in microbial popula-

tions: insights from theoretical, comparative and experimental studies.Heredity 100:471–477.

Marriott, A. C., and N. J. Dimmock. 2010. Defective interfering viruses andtheir potential as antiviral agents. Rev. Med. Virol. 20:51–62.

May, R. M., and M. A. Nowak. 1995. Coinfection and the evolution of parasitevirulence. Proc. R. Soc. Lond. B 261:209–215.

Meynell, E., and N. Datta. 1967. Mutant drug resistant factors of high trans-missibility. Nature 214:885–887.

Modi, R. A., and J. Adams. 1991. Coevolution in bacterial-plasmid popula-tions. Evolution 45:656–667.

Nowak, M. A., and R. M. May. 1994. Superinfection and the evolution ofparasite virulence. Proc. R. Soc. Lond. B 255:81–89.

Ou, J. T. 1980. Role of surface exclusion genes in lethal zygosis in Escherichia

coli K12 mating. Mol. Gen. Genet. 178:573–581.R Core Development Team. 2011. R: a language and environment for statistical

computing. R Foundation for Statistical Computing, Vienna, Austria.

Rankin, D. J., K. Bargum, and H. Kokko. 2007. The tragedy of the commonsin evolutionary biology. Trends Ecol. Evol. 22:643–651.

Read, A. F., M. J. Mackinnon, M. A. Anwar, and L. H. Taylor. 2002. Kin-selection models as evolutionary explanations of malaria. Pp. 165–178in U. Dieckmann, J. A. J. Metz, M. W. Sabelis, and K. Sigmund, eds.Adaptive dynamics of infectious diseases: in pursuit of virulence man-agement. Cambridge Univ. Press, Cambridge.

Read, A. F., and L. H. Taylor. 2001. The ecology of genetically diverseinfections. Science 292:1099–1102.

Rumbaugh, K. P., S. P. Diggle, C. M. Watters, A. Ross-Gillespie, A. S. Griffin,and S. A. West. 2009. Quorum sensing and the social evolution ofbacterial virulence. Curr. Biol. 19:341–345.

Smith, A. 1776. An inquiry into the Nature And Causes of the Wealth ofNations. W. Strahan and T. Cadell, London.

Smith, J. 2001. The social evolution of bacterial pathogenesis. Proc. R. Soc.Lond. B 268:61–69.

smith, J. 2011. Superinfection drives virulence evolution in experimental pop-ulations of bacteria and plasmids. Evolution 65:831–841.

Su, L.-H., C. Chu, A. Cloeckaert, and C.-H. Chiu. 2008. An epidemic of plas-mids? Dissemination of extended-spectrum cephalosporinases amongSalmonella and other Enterobacteriaceae. FEMS Immun. Med. Micro-biol. 52:155–168.

Turner, P. E., V. S. Cooper, and R. E. Lenski. 1998. Tradeoff between horizon-tal and vertical modes of transmission in bacterial plasmids. Evolution52:315–329.

West, S. A., S. P. Diggle, A. Buckling, A. Gardner, and A. S. Griffin.2007. The social lives of microbes. Annu. Rev. Ecol. Evol. Syst. 38:53–77.

Associate Editor: S. Gandon

Supporting InformationThe following supporting information is available for this article:

Figure S1. Population dynamics expected in serial transfer assay of plasmid fitness.

Table S1. Parameter values used in numerical solutions.

Supporting Information may be found in the online version of this article.

Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting information supplied by the

authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

1 2 7 4 EVOLUTION APRIL 2012