Phenotypic Plasticity in Bacterial Plasmids

Corresponding author: Paul E. Turner, Science Area Receiving, 266 Whitney Ave., Yale University, New Haven, CT 06511., paul.turner{at}yale.edu (E-mail)

Communicating editor: H. OCHMAN

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Plasmid pB15 was previously shown to evolve increased horizontal (infectious) transfer at the expense of reduced vertical (intergenerational) transfer and vice versa, a key trade-off assumed in theories of parasite virulence. Whereas the models predict that susceptible host abundance should determine which mode of transfer is selectively favored, host density failed to mediate the trade-off in pB15. One possibility is that the plasmid's transfer deviates from the assumption that horizontal spread (conjugation) occurs in direct proportion to cell density. I tested this hypothesis using Escherichia coli/pB15 associations in laboratory serial culture. Contrary to most models of plasmid transfer kinetics, my data show that pB15 invades static (nonshaking) bacterial cultures only at intermediate densities. The results can be explained by phenotypic plasticity in traits governing plasmid transfer. As cells become more numerous, the plasmid's conjugative transfer unexpectedly declines, while the trade-off between transmission routes causes vertical transfer to increase. Thus, at intermediate densities the plasmid's horizontal transfer can offset selection against plasmid-bearing cells, but at high densities pB15 conjugates so poorly that it cannot invade. I discuss adaptive vs. nonadaptive causes for the phenotypic plasticity, as well as potential mechanisms that may lead to complex transfer dynamics of plasmids in liquid environments.

ALTHOUGH plasmids can provide beneficial traits such as antibiotic resistance, in the absence of positive selection a plasmid typically reduces the growth of its bacterial host in vitro (LEVIN 1980 ; ZUND and LEBEK 1980 ; LENSKI and BOUMA 1987 ; NGUYEN et al. 1989 ) and in situ (DEVANAS et al. 1986 ; SUNDIN et al. 1994 ; BJORKLOF et al. 1995 ; MOENNE-LOCCOZ and WEAVER 1995 ). Plasmids are vertically transferred to daughter cells during binary fission and can be horizontally transferred from infected cells (donors) to uninfected cells (recipients) through conjugation (LEDERBERG and TATUM 1946 ; CLEWELL 1993 ; THOMAS 2000 ). Plasmid activities that promote horizontal transmission (such as producing more conjugative pili) should further impede host growth, thereby reducing the plasmid's potential for vertical transmission. We previously demonstrated this trade-off when plasmids evolved increased horizontal transfer at the expense of decreased vertical transfer and vice versa (TURNER et al. 1998 ).

Spread of conjugative plasmids in liquid culture is often approximated through simple mass-action models (STEWART and LEVIN 1977 ; LEVIN et al. 1979 ; BERGSTROM et al. 2000 ; GANUSOV and BRILKOV 2002 ; PAULSSON 2002 ). Matings between donors and recipients are assumed to be governed by random encounters, which result in the creation of plasmid-bearing transconjugants. If matings occur in proportion to donor and recipient concentrations the rate of change of plasmid-bearing cells is

where P is the combined densities (cells per milliliter) of donors and transconjugants, R is recipient density, _P is exponential growth rate of plasmid-bearing cells, and is the rate of conjugative transmission. The per-capita rate,

reveals that vertical spread (_P) is independent of host density, whereas horizontal spread (R) should be proportional to recipient density.

For parasites that feature both horizontal and vertical modes of transmission, susceptible host density (R) should determine which mode is selectively favored (MAY and ANDERSON 1983 ; HERRE 1993 ; BULL 1994 ; EWALD 1994 ), provided there is a genetic trade-off between transfer modes. Consider three hypothetical plasmids: x, y, and z. Plasmid x conjugates faster than plasmid y, but the trade-off causes slower growth of x-infected hosts. Plasmid z transfers only vertically, thereby causing the least host burden. The model predicts that x and y should spread through the host population faster as R increases (LEVIN et al. 1979 ; SIMONSEN et al. 1990 ; Fig 1A). In this way, a costly plasmid can theoretically invade a population if its rate of horizontal transfer is rapid enough to offset its poor vertical transmission (STEWART and LEVIN 1977 ). At high densities x is most favored to invade, whereas the advantage progressively shifts to plasmids y and z as R decreases. However, any plasmid can fail to invade if its net rate of spread is negative (r < _R – P), where _R is the growth rate of recipients (Fig 1A).

View larger version (18K): In this window In a new window Download PPT slide   Figure 1. Theorized effect of susceptible host density on plasmid transmission and phenotypic plasticity of plasmid traits. (A) Hypothetical net rates of increase (solid lines) for plasmid genotypes x, y, and z, as a function of available host density. The rapidly conjugating plasmid x is favored at high densities, but a trade-off between transmission modes causes the advantage to shift in favor of plasmids y and z as densities decline. For any plasmid to spread, the net rate must exceed the per-capita growth rate of plasmid-free recipients in the population (dashed line). (B) Reaction norms for two hypothetical plasmids featuring traits unaffected by host density. The solid line designates a plasmid that conjugates slowly but imposes a low cost of carriage, whereas the dashed line indicates a faster-conjugating plasmid that imposes a higher cost. (C) Reaction norms for two hypothetical plasmids featuring phenotypically plastic traits. The solid line shows a plasmid whose conjugation rate and cost of carriage decline with increasing host density, whereas the dashed line symbolizes a plasmid featuring the opposite response. Most models predict that rapid conjugation can be selectively favored at high densities; thus, a correspondingly "optimal" plasmid should feature the latter form of plasticity.

Plasmid transfer on surfaces (such as in biofilms) is more complicated than in well-agitated liquid culture, because heterogeneous environments can lead to "patchy" distribution of donors and recipients (SIMONSEN 1990 ; LICHT et al. 1999 ; LAGIDO et al. 2003 ). However, in either environment increasing cell densities should generally improve plasmid spread (up to a possible saturation point) because the greater cell-to-cell contact affords more opportunities for horizontal transfer. The improvement is generally not due to changes in the rate constant of conjugative transfer () or in the cost of plasmid carriage (c), which can be estimated as the difference between recipient and donor growth rates:

[Cost of carriage can be more broadly defined in terms of the fitness difference between plasmid-free and plasmid-bearing hosts (BOUMA and LENSKI 1988 ; TURNER et al. 1998 ), whereby growth rate is only one of several fitness components.] Thus, both and c are assumed to remain constant as host density varies (Fig 1B). But in the absence of genetic constraints, selection should favor parasites that are phenotypically plastic and able to switch between transmission strategies in direct response to environmental conditions (BRADSHAW 1965 ; SCHEINER 1993 ). Therefore, a phenotypically plastic plasmid might be expected to maximize the conjugation rate at high densities to facilitate horizontal spread and to minimize conjugation at low densities to promote vertical spread (Fig 1C).

In a previous study (TURNER et al. 1998 ) we predicted that high host densities should favor evolution of increased plasmid conjugation at the expense of reduced vertical transmission, whereas low densities should favor the reverse. Plasmid pB15 (LUNDQUIST and LEVIN 1986 ) was allowed to evolve for 500 generations in replicated batch-culture environments containing different availabilities of uninfected Escherichia coli hosts. Bacterial concentrations (governed by sugar concentration in the liquid medium) were high (10⁹ cells/ml) to facilitate contact between donors and recipients, but differences in available host density were achieved by manipulating the immigration of recipients into treatment populations. To prevent disruption of mating pairs the experimental populations were grown in static culture, where matings can also occur between a minority subpopulation of cells at the bottom surface of the culture tube. Results showed that plasmids evolved greater conjugation at the expense of reduced vertical transfer (and vice versa), confirming the implicit trade-off described in many models for the evolution of parasite virulence (damage to host fitness; MAY and ANDERSON 1983 ; BULL 1994 ; EWALD 1994 ). However, available host density did not mediate the trade-off because plasmids featuring increased (and reduced) conjugation evolved in all treatments (TURNER et al. 1998 ).

Failure of susceptible host density to mediate the trade-off in pB15 can be explained if plasmid spread does not increase in proportion to cell concentration. Here I examine the effects of cell density on horizontal transfer of pB15 and demonstrate that invasiveness decreases with cell concentration in static culture environments. In addition, I show the phenomenon is due to phenotypic plasticity in traits governing vertical and horizontal transfer of the plasmid. At low cell densities pB15 maximizes horizontal transmission at the expense of vertical transfer, but at high cell densities it features the reverse.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Strains:
Table 1 lists strains used in this study. Hosts (kindly provided by R. Lenski, Michigan State University, East Lansing, MI) were derived from a single clone of E. coli B (REL1206), which evolved previously for 2000 generations in a glucose-limited environment (LENSKI et al. 1991 ); this strain cannot grow on L-arabinose and is denoted Ara^–. Ara⁺ and Ara^– strains form white and red colonies, respectively, on tetrazolium-arabinose (TA) indicator plates (LEVIN et al. 1977 ). The Ara marker is selectively neutral under the shaking and static culture conditions used here (TRAVISANO 1997 ; TURNER et al. 1998 ).

Plasmids were obtained from the laboratory of B. Levin (Emory University, Atlanta, GA). R1 is a large (100 kb) well-described plasmid of the IncFII group, featuring a copy number of four to five copies per cell (NORDSTROM et al. 1980 ). R1 (and many other plasmids) normally repress conjugative pilus synthesis, but upon transfer to a new cell the plasmid's transfer operon is transitorily derepressed for a short time to facilitate further transmission (WILLETTS 1974 ; LUNDQUIST and LEVIN 1986 ). In contrast, plasmid R1-drd19 used in this study is a mutant of R1 that is permanently derepressed for conjugation. R1-drd19 confers clinical resistance to kanamycin (Km) and several other antibiotics (LUNDQUIST and LEVIN 1986 ). Plasmid pB15 is also large (50 kb), but its copy number per cell is unknown. Although pB15 is not fully characterized, preliminary sequence data suggest it is related to R64, an IncI1 plasmid of Salmonella (D. GUTTMAN and P. TURNER, unpublished results). Plasmid pB15 conjugates at high rates in chemostats containing 10⁸ cells/ml in 50 µg/ml glucose medium (LUNDQUIST and LEVIN 1986 ). It confers clinical resistance to Km and subclinical resistance to tetracycline (Tc; LUNDQUIST and LEVIN 1986 ), confirmed by growth on TA with 25 µg/ml Km and 1 µg/ml Tc, respectively. To move a plasmid into REL1206, this recipient was mixed with plasmid-bearing donors, and transconjugants were obtained through overnight matings. No differences in plating efficiency on selective and nonselective media were observed for any of the plasmid-bearing cells in this study.

Plasmid segregation is incomplete transfer to one of the daughter cells during binary fission. R1-drd19 features at least one mechanism to ensure its stability in the absence of selection for plasmid carriage (GERDES et al. 1986 ). It is unknown whether pB15 features a stability mechanism such as postsegregational killing (HELINSKI et al. 1996 ), but pB15 segregants are fairly rare; for strains REL5382 and REL5384 (Table 1), 1 in 4000 colonies formed on TA is a segregant on the basis of tooth-picking assays onto TA + Km (TURNER et al. 1998 ). Thus, both pB15 and R1-drd19 can be considered relatively stable under the current culture conditions.

Culture conditions:
Bacteria were grown at 37° in batch culture using Davis minimal (DM) broth (CARLTON and BROWN 1981 ) supplemented with 2 µg/ml thiamine hydrochloride and glucose (glu) at a specific concentration; e.g., DM25 indicates DM broth with 25 µg/ml glu, which yields 5 x 10⁷ cells/ml at stationary phase. Culture volume was 10 ml, in nonshaking 18 x 150-mm glass tubes or shaking 50-ml Erlenmeyer flasks. Daily propagation occurred by vortexing a culture, followed by 100-fold dilution into fresh medium (serial transfer). During this 24-hr cycle, bacteria attained stationary-phase densities, at which point they had depleted the available resource. The resulting 100-fold growth of the population represents 6.64 (= log₂ 100) generations of binary fission per day.

Invasion-when-rare experiments:
To examine plasmid spread, donors and recipients were mixed 1:200 in DM broth containing 10, 12.5, 25, 50, 100, 200, 400, 800, or 1000 µg/ml glu. Mixtures were serially transferred for up to 20 days, in the presence or absence of shaking. Every day, after serial transfer had taken place, glycerol was added to each population, which was then stored in a freezer at –80° for future study. Daily samples were plated on TA to track recipient densities and on TA with 25 µg/ml Km to measure densities of donors and transconjugants. (Transconjugants were also tested on TA with 1 µg/ml Tc and no dissociation between resistance markers was observed, indicating that plasmids did not lose the Km marker over time as they were transferred between cells.) In some experiments, populations were sampled on minimal-arabinose (MA) plates containing Km, to screen for Ara⁺ transconjugants that were very rare relative to Ara^– donors. Km is an aminoglycoside that is bactericidal to sensitive cells, as confirmed by spreading plasmid-free recipients onto TA + Km. Therefore, transconjugant estimates were not confounded by matings between recipients and plasmid-bearing cells on TA + Km plates. To further dismiss plate matings as a potential confounding factor, diluted samples from frozen populations were spread on both TA + Km and MA + Km plates; colony counts yielded identical estimates of transconjugant densities per milliliter.

Conjugation rate assay:
To assay conjugation rate (), Ara^–/pB15 donors and Ara⁺ Nal^r (nalidixic acid resistant) recipients were mixed 1:100 and allowed to grow and mate during a standard 24-hr growth cycle in static culture. The Nal^r marker facilitated visualization of rare transconjugants on selective plates containing 25 µg/ml Km and 15 µg/ml Nal. After 24 hr, the final densities of donors (D), recipients (R), and transconjugants (T) were determined by colonies formed on selective and nonselective plates. Growth rate per hour in exponential phase () of mating cultures was estimated by regressing the natural logarithm of total cell density vs. time during the period of exponential-phase growth. The rate of conjugative transfer (milliliters per cell hour) for matings in batch culture may be estimated using the formula

(SIMONSEN et al. 1990 ), where N = T + R + D and N₀ is initial population size. Unlike other transfer measures (e.g., WATANABE 1963 ; BALE et al. 1987 ), the end-point method is largely unaffected by factors such as donor-to-recipient ratio because it estimates the actual rate constant of transfer rather than simply the resulting frequency of transconjugants (SIMONSEN et al. 1990 ). The equation is simplified by using for the total population, thereby ignoring predictably slower growth of plasmid-bearing cells. However, even reasonably large costs of carriage do not strongly affect estimation of (SIMONSEN et al. 1990 ).

Relative fitness and cost of plasmid carriage:
To estimate relative fitness, two strains (distinguished by the Ara marker) were competed under the culture regimes described above. Strains were grown separately (preconditioned) for 1 day in the experimental medium to ensure comparable physiological states. They were then mixed 1:1, diluted 1:100 into fresh medium, and allowed to grow and compete for 24 hr. Initial and final densities of each competitor were estimated on TA plates.

Let the initial densities of the two competitors be N₁(0) and N₂(0), respectively, and let N₁(1) and N₂(1) be their densities after 1 day. The time-average rate of increase, m_i, for each competitor was then calculated as

The fitness of one strain relative to the other is expressed simply as the dimensionless ratio of their rates of increase:

(LENSKI et al. 1991 ). Cost of plasmid carriage was measured by competing a plasmid-bearing strain against its plasmid-free counterpart that differed by the neutral Ara marker. Fitness of the plasmid-bearing strain relative to the plasmid-free strain was calculated as above. Cost of carriage (c) is the difference between 1.0 and the estimated fitness; thus, c > 0 indicates that a plasmid reduces host fitness. Fitness estimates are complicated, in principle, by segregants and transconjugants that arise during competitions between plasmid-bearing and plasmid-free cells. To ensure accurate estimates, I sampled a subset of cells to track plasmid losses and gains due to segregation and conjugation that occurred during competitions. As previously observed (TURNER et al. 1998 ), segregants and transconjugants were small minorities, and their inclusion or exclusion from the calculations had no significant effect on fitness estimates.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Invasion by pB15 is maximal at intermediate densities in static culture:
Invasion-when-rare experiments in static environments were used to examine the effects of cell density on spread of plasmid pB15. In these assays the minority donors (Ara⁺/pB15) should decline due to the cost of plasmid carriage, whereas the majority recipients (Ara^–) should remain roughly constant (unless Ara^–/pB15 transconjugants become very numerous). These predictions were generally supported, confirming that pB15 was costly in all treatments (Fig 2). In some cases, donors decreased to low densities where (presumably) their loss due to selection was balanced by their increase through reinfection of rare segregants.

View larger version (30K): In this window In a new window Download PPT slide   Figure 2. Invasion by plasmid pB15 occurs only at intermediate host densities in static culture. Log10 colony-forming units (CFU) per milliliter of Ara– recipients (circles), Ara+/pB15 donors (squares), and Ara–/pB15 transconjugants (diamonds) during serial transfer in seven environments are shown: (A) 12.5, (B) 25, (C) 50, (D) 100, (E) 200, (F) 400, and (G) 800 µg/ml glucose.

Invasion success was gauged by tracking transconjugant densities. Results (Fig 2) showed that the rate of transconjugant formation per day was not sufficient to outpace selection against plasmid carriage at 12.5 glu (linear regression with slope = –0.1408, t = 1.886, d.f. = 4, P = 0.132) and 25 glu (slope = –0.0789, t = 3.452, d.f. = 7, P = 0.011). (Assay at 12.5 glu was halted at 11 days due to contamination, but by this time transconjugants were already below the limit of detection on TA + Km.) Consistent with the predicted effect of increased cell density, pB15 invasion improved at intermediate glucose concentrations: 50 glu (slope = 0.0979, t = 13.629, d.f. = 14, P < 0.0001) and 100 glu (slope = 0.0944, t = 14.139, d.f. = 14, P < 0.0001). But at high concentrations, the plasmid invaded poorly: 200 glu (slope = –0.0173, t = 5.405, d.f. = 15, P < 0.0001), 400 glu (slope = –0.1287, t = 7.555, d.f. = 15, P < 0.0001), and 800 glu (slope = –0.0888, t = 3.523, d.f. = 14, P = 0.003). Transconjugants disappeared faster than donors in the latter two environments. This result could occur if the cost of carriage is higher in Ara^– cells, but this idea seems unlikely given marker neutrality under the current conditions (TRAVISANO 1997 ; TURNER et al. 1998 ). A plausible explanation is that the Ara⁺ donors are more numerous than the Ara^– transconjugants, causing the Ara⁺ background to provide a larger segregant pool for reinfection. There was also a sharp drop in donor and transconjugant densities on day 12 in 400 glu. This decline could easily result from a pipette error causing >100-fold dilution during serial passage; the plasmid-bearing cells were in the minority and sampling error could cause them to be underrepresented in the propagated cells.

Fig 3A shows the data for rates of transconjugant formation (log₁₀ transconjugants/ml/day) vs. mean log₁₀ R (stationary-phase recipient density). Clearly, the invasion rate of pB15 is positive (and maximal) only at the intermediate resource concentrations. My results demonstrate that spread of pB15 in static culture does not increase in direct proportion to cell density, as governed by glucose concentration.

View larger version (17K): In this window In a new window Download PPT slide   Figure 3. Invasion by plasmid pB15 is maximal at intermediate host densities in static culture. (A) Solid circles summarize the experiments shown in Fig 2. Solid triangles are from a second set of invasion assays in fewer environments (25, 100, and 800 µg/ml glucose), using Ara+ recipients and Ara–/pB15 donors. The solid line is a quadratic fit to the data. (B) Seven independent invasion assays (12.5, 25, 50, 100, 200, 400, and 800 µg/ml glu), using Ara–-marked donor and recipient cells (solid squares, dashed line) and using Ara+-marked donors and recipients (solid diamonds, solid line). See text for details.

Invasion by pB15 is unaffected by chromosomal markers:
The above results could be due to an unexpected interaction between pB15 and the Ara^– marker on the recipient chromosome. To examine this possibility, two experiments were conducted. First, Ara^–/pB15 donors invaded Ara⁺ recipients for 9 days (sufficient time to verify previous dynamics) at three glucose concentrations. In these assays recipient densities remained approximately constant, whereas donors declined due to plasmid carriage (data not shown). More importantly, the rate of transconjugant formation per day (Fig 3A) was qualitatively similar to that of the above assays involving the opposite marker combination for donors and recipients: 25 glu (slope = –0.1258, d.f. = 7, P = 0.009), 100 glu (slope = 0.1461, d.f. = 7, P = 0.001), and 800 glu (slope = –0.1301, d.f. = 7, P = 0.002).

Second, invasion experiments were repeated at seven glucose concentrations for 10 days, but donors and recipients featured the identical Ara marker. These assays are less accurate because donors cannot be distinguished from transconjugants, but they can determine whether pB15 transfer between differently marked cells accounts for poor invasion at high cell densities. Consistency between these and the above assays would be maximal plasmid spread at intermediate glucose concentrations. However, the maximum rate can be net negative because majority donors can decline as minority transconjugants increase. Once again, results showed that the rate of plasmid spread was most rapid at intermediate glucose concentrations, regardless of the Ara marker shared by donors and recipients (Fig 3B).

Invasion by R1-drd19 is maximal at high densities in static culture:
Failure of pB15 to invade fastest at high cell densities could be due to particulars of the culture regime. To examine this potential bias, I conducted invasion experiments using a different plasmid. Ara^–/R1-drd19 donors invaded Ara⁺ recipients for 5 days at three glucose concentrations, with threefold replication. Results (Fig 4) showed that the recipients remained approximately constant, whereas the donors declined due to plasmid carriage. As cell density increased, the mean rate of change in transconjugants also increased: 10 glu (slope = –0.9283, d.f. = 1, P = 0.181), 100 glu (slope = 0.1742, d.f. = 3, P = 0.001), and 1000 glu (slope = 0.6299, d.f. = 3, P = 0.0018). These data showed that the static culture regime did not bias against increased plasmid spread at higher glucose concentrations.

View larger version (16K): In this window In a new window Download PPT slide   Figure 4. Invasion by plasmid R1-drd19 occurs in proportion to host density in static culture. Each point represents the mean of independent assays (n = 3). Log10 CFU/ml of Ara+ recipients (circles), Ara–/R1-drd19 donors (squares), and Ara+/R1-drd19 transconjugants (diamonds) during serial transfer in three environments are shown: (A) 10, (B) 100, and (C) 1000 µg/ml glucose.

Mass action governs invasion by R1-drd19 and pB15 in shaking environments:
To examine whether mass action governs plasmid spread in shaking environments, I conducted invasions using R1-drd19 and pB15 in shaking culture. Ara^–/R1-drd19 donors invaded Ara⁺ recipients, whereas Ara⁺/pB15 donors invaded Ara^– recipients. Both mating combinations were replicated threefold, at 100 and 1000 glu. Results (Fig 5) showed that for R1-drd19 the mean rate of change in transconjugants per day was faster at high glucose concentration, similar to the outcome in static culture: 100 glu (slope = 0.5578, d.f. = 3, P = 0.024) and 1000 glu (slope = 0.6179, d.f. = 3, P = 0.017). More importantly, the data revealed that pB15 can successfully invade at intermediate and high densities in shaking environments: 100 glu (slope = 0.7921, d.f. = 2, P = 0.031) and 1000 glu (slope = 0.8264, d.f. = 3, P = 0.0048). These results indicated that mass action is a reasonable descriptor of pB15 spread in well-agitated liquid culture.

View larger version (23K): In this window In a new window Download PPT slide   Figure 5. Invasions by plasmids R1-drd19 and pB15 occur in proportion to host density in shaking culture. Each point represents the mean (±SE) of independent assays (n = 3); error bars too small to be visualized are omitted. (A and B) Log10 CFU/ml of Ara+ recipients (circles), Ara–/R1-drd19 donors (squares), and Ara+/R1-drd19 transconjugants (diamonds) during serial transfer in two environments: (A) 100 and (B) 1000 µg/ml glucose. (C and D) Identical experiments using Ara– recipients (circles), Ara+/pB15 donors (squares), and Ara–/pB15 transconjugants (diamonds): (C) 100 and (D) 1000 µg/ml glucose.

Conjugation rate of pB15 declines at high densities in static environments:
The horizontal component of plasmid spread (R, see Introduction) is assumed to be proportional to the density of potential recipients (R). Because this prediction for pB15 breaks down in static culture (Fig 3), it suggests that conjugation rate () or host density (R) (or both) deviates from expectations.

I first tested whether glucose concentration predictably governs R in static culture. To do so, I calculated mean log₁₀ R at each glucose concentration using the data from pB15 invasions in static environments (Fig 2): 12.5 glu, 3.24 x 10⁷ cells/ml; 25 glu, 5.88 x 10⁷ cells/ml; 50 glu, 8.95 x 10⁷ cells/ml; 100 glu, 1.67 x 10⁸ cells/ml; 200 glu, 3.45 x 10⁸ cells/ml; 400 glu, 6.30 x 10⁸ cells/ml; and 800 glu, 9.90 x 10⁸ cells/ml. This analysis showed that cell density approximately doubled as the glucose concentration similarly increased, and a linear regression of mean log₁₀ R on glucose concentration was statistically significant (slope is 0.0017, d.f. = 5, t = 4.046, P = 0.0099).

To examine whether pB15 conjugation is constant across cell densities, I measured at 5, 50, 500, and 1000 glu in static culture; six blocks of assays were performed. Results showed that was highly sensitive to cell density governed by glucose concentration (Fig 6A). A two-way ANOVA showed a highly significant effect of glu on , but no block effect (Table 2). Mean (n = 6) for pB15 at 5 glu was 1.092 x 10^–10 ml/cell hour; that is, during 1 hr, each plasmid-bearing cell can effectively "search" a volume of 10^–10 ml and infect any plasmid-free cell therein. (Because cell densities are <<1/, a donor is unlikely to encounter two recipients in close temporal proximity, or vice versa; hence, the system is unsaturated.) In contrast, mean progressively diminished at 50 glu (5.315 x 10^–11 ml/cell hour), 500 glu (6.185 x 10^–12 ml/cell hour), and 1000 glu (9.483 x 10^–13 ml/cell hour). To further examine this phenomenon, I estimated mean R (n = 6) at each glucose concentration: 5 glu, 1.65 x 10⁷ cells/ml; 50 glu, 8.81 x 10⁷ cells/ml; 500 glu, 8.76 x 10⁸ cells/ml; and 1000 glu, 2.16 x 10⁹ cells/ml. Using mean values of and R, it is revealed that R in pB15 does not increase with glucose concentration. Rather, R is approximately constant between low and intermediate cell densities (5 glu, 0.00542; 50 glu, 0.00533) and even declines slightly at the highest experimental densities (500 glu, 0.00180; 1000 glu, 0.00205). This phenomenon occurs because is not constant, but declines by approximately an order of magnitude as R increases by the same amount.

View larger version (15K): In this window In a new window Download PPT slide   Figure 6. Plasmid pB15 features phenotypically plastic traits in static culture environments. Each point represents an independent measurement. (A) Conjugation rate (solid circles) of pB15 decreases with increasing cell concentration (Log10 CFU/ml). See Table 2 for statistics. (B) Cost of plasmid carriage (solid triangles) similarly declines, due to the trade-off between horizontal and vertical modes of transmission. See text for statistical analyses.

View this table: In this window In a new window   Table 2. ANOVA of the effect of glucose concentration on conjugation rate () of plasmid pB15

Trade-off between vertical and horizontal transfer of pB15 holds across cell densities:
Because vertical and horizontal transfer routes in pB15 show a trade-off (TURNER et al. 1998 ), the cost of plasmid carriage should vary along with across glucose concentrations. To test this hypothesis, I measured the fitness of Ara⁺/pB15 cells relative to Ara^– cells in seven glucose concentrations (12.5, 25, 50, 100, 200, 400, 800, and 1000 glu) in static culture, with replication (n = 2). I then calculated the cost of plasmid carriage, c (see MATERIALS AND METHODS). Results (Fig 6B) indicated that plasmid-bearing cells were generally disadvantaged, as all values of c exceeded zero. More importantly, these data showed that c decreased at higher glucose concentrations, and this outcome was statistically significant (linear regression with slope = –0.1341, d.f. = 14, t = 4.487, P = 0.0005). These data demonstrate that the genetic trade-off between transmission modes in pB15 holds across a wide range of cell densities and that both and c are phenotypically plastic in static culture containing glucose.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Many parasites can transfer vertically between parent and offspring, as well as horizontally between infected and uninfected individuals. Activities that augment horizontal transmission (such as greater within-host reproduction) are assumed to reduce host fitness, thereby decreasing vertical transmission (MAY and ANDERSON 1983 ; HERRE 1993 ; BULL 1994 ; EWALD 1994 ; but see EBERT and BULL 2003 ). We previously observed that plasmid pB15 evolves greater conjugation at the expense of reduced intergenerational transfer (and vice versa) in static culture (TURNER et al. 1998 ), demonstrating the trade-off between transmission modes commonly assumed in theories for the evolution of parasite virulence (see also DIFFLEY et al. 1987 ; DEARSLEY et al. 1990 ; BULL et al. 1991 ; EBERT and MANGIN 1997 ; MESSENGER et al. 1999 ). Although the models predict that susceptible host abundance should determine which transfer mode is selectively favored, host density failed to mediate the trade-off in pB15 because plasmids evolved increased (and reduced) conjugation in all treatments (TURNER et al. 1998 ).

Theory generally predicts that plasmid spread should improve with increasing cell density due to greater contact between donor and recipient cells (e.g., STEWART and LEVIN 1977 ; LEVIN et al. 1979 ; SIMONSEN 1990 ; LAGIDO et al. 2003 ). Here I demonstrated that host density failed to mediate the trade-off in pB15 because the plasmid's rate of horizontal spread is not proportional to cell density in static culture. Rather, invasion by pB15 is maximal at intermediate cell densities determined by glucose concentration. This result explains why manipulation of available host density at elevated glucose concentrations (i.e., 10⁹ cells/ml, 1000 µg/ml glu) did not predictably select for the evolution of conjugation rates in pB15-derived plasmids (TURNER et al. 1998 ; see further discussion below).

Spread of pB15 occurs disproportionately to cell density in static culture because traits governing its transfer are phenotypically plastic. The rate of conjugative transfer () is expected to be a constant, which is relatively insensitive to cell density, donor-to-recipient ratio, and other environmental factors (SIMONSEN et al. 1990 ). Thus, horizontal spread should be determined by the product of and susceptible host density (R, a variable determined by resource concentration), causing R to increase in direct proportion to R. In contrast, R in pB15 is relatively unchanged as R increases, because declines by approximately an order of magnitude as R is similarly increased in static culture containing glucose. Although it is clear that increased cell densities negatively impact spread of pB15 in static culture, it is questionable whether the mass-action measurement should be used to describe the phenomenon. SIMONSEN 1990 showed that measurements of are strikingly similar for plasmid R1-drd19 when bacteria are mated on surfaces and in shaking liquid culture, even when cells are grown at densities below those used in my experiments (i.e., where donors very rarely encounter recipients). Therefore, the data by SIMONSEN 1990 suggest that may be usefully applied in static culture environments where matings occur both in liquid and between cells that settle out in culture. However, non-mass-action models should more accurately describe conjugal transfer in complex environments such as on surfaces (e.g., LAGIDO et al. 2003 ).

At intermediate densities horizontal transfer in pB15 is sufficient to overcome the growth disadvantage suffered by plasmid-bearing cells, but at high densities its poor conjugation prevents the plasmid from invading. My data allow a crude estimate of the cell densities permitting successful invasion. For pB15, the rate of change in transconjugants per milliliter per day is apparently a nonlinear function of log₁₀ R (Fig 3A), and a quadratic function provides a statistically significant fit to the data (F_[2,7] = 7.0428, P = 0.0211). This function is described by dT/dt = –29.957 + 7.268 x R – 0.439 x R², which can be solved for the two values of R where the curvilinear fit crosses a value of zero. This solution yields R = 7.244 x 10⁷ cells/ml and 4.571 x 10⁸ cells/ml as the lower and upper boundaries, respectively, for successful invasion by pB15 in static culture. Although the estimate indicates a relatively narrow range of existence conditions for pB15, more thorough examination of invasion conditions can be explored using the parameters described in this study.

The cost of plasmid carriage in pB15 is also phenotypically plastic in static culture, because the fitness disadvantage suffered by plasmid-bearing cells declines as hosts become more numerous. Thus, the previously described trade-off between transmission modes (TURNER et al. 1998 ) is now revealed to hold across a broad range of cell concentrations in static culture. However, the plasticity observed in pB15 seems nonintuitive because it indicates that the plasmid relies more heavily upon vertical transfer as cell densities are increased.

Adaptive vs. nonadaptive phenotypic plasticity:
Phenotypic plasticity may or may not be the result of adaptation (see SCHEINER 1993 for review). If plasmid transfer is proportional to cell density, it seems logical that "optimal" phenotypic plasticity should allow increased plasmid virulence (more rapid conjugation) at high host densities and decreased virulence (reduced conjugation) at low host densities. Precisely the opposite was observed in pB15, suggesting the plasticity is nonadaptive. However, the theorized importance of cell density for plasmid spread is a reasonably good predictor for invasiveness of plasmid R1-drd19 in static and shaking environments and of pB15 in shaking culture. These combined results strongly suggest that an interaction between pB15 and the static culture environment governs the plasticity observed.

One possibility is that the plasticity is a coincidence of the novel association between plasmid pB15 and E. coli B bacteria. The experimental host was evolved in vitro for 2000 generations (LENSKI et al. 1991 ), whereas pB15 was isolated from an unknown E. coli host taken from a human undergoing antibiotic treatment (LUNDQUIST and LEVIN 1986 ). It may be that an epistatic interaction between plasmid and chromosomal genes is responsible in part (or wholly) for the observed plasticity. To examine this hypothesis, one could look for phenotypic plasticity when the plasmid is moved to a different host background, such as E. coli K-12 or unevolved E. coli B. Equally intriguing would be to examine the "cost of plasticity" in pB15, by predicting that evolution of plasmid/host associations in a constant environment would select against maintenance of phenotypic plasticity, assuming there is a cost of maintaining the associated genetic machinery; Drosophila experiments demonstrate that plasticity of a trait can change in response to selection (see SCHEINER 1993 for review). This idea could be studied by evolving E. coli B/pB15 associations for prolonged periods in static culture at a single glucose concentration. In fact, this experiment has already been completed in the control populations from TURNER et al. 1998 , where replicated bacteria/plasmid associations were allowed to evolve for 500 generations in 1000 µg/ml glu. Thus, evolved plasmids from the prior study could be examined for changes in plasticity.

A second possibility is that the phenotypic plasticity is a consequence of differing physiology of the E. coli host at varying glucose (and hence host density) levels. The 2000 generations of evolution experienced by the bacterium at 25 µg/ml glu in shaking culture (LENSKI et al. 1991 ) are sufficient for profound adaptive changes, which can cause very different performance in other environments (e.g., TRAVISANO 1997 ). Therefore, it is possible that the host's physiology could be greatly affected by growth in elevated glucose concentrations and/or static culture. In turn, because plasmids are obligate intracellular elements, expression of plasmid genes (such as conjugative ability) can be strongly affected by the host's physiological state (LEVIN et al. 1979 ). This idea is supported by the phenomenon of transitory derepression. Immediately after conjugation has occurred, the mechanism allows a plasmid to temporarily derepress its conjugation to maximize the rate of further horizontal transfer, suggesting transitory derepression occurs in response to host physiology.

At this juncture, it appears that pB15's conjugation rate is strongly influenced by levels of cell density in static culture. Preliminary sequence data for pB15 (D. GUTTMAN and P. TURNER, unpublished data) suggest that the plasmid may be related to R64, a large (120 kb) IncI1 plasmid of Salmonella typhimurium (HEDGES and DATTA 1973 ). IncI1 plasmids form two types of sex pili, a thin flexible pilus and a thick rigid pilus (BRADLEY 1983 , BRADLEY 1984 ). The thick pilus is essential for mating in general, whereas the thin pilus is required only for liquid mating. R64 features an elaborate 54-kb transfer mechanism, the R64 shufflon, which is prone to stochastic DNA rearrangements that strongly influence mating efficiency, especially recipient specificity in liquid culture (KOMANO et al. 1995 ; KOMANO 1999 ). Although pB15 appears genetically similar to R64, its transfer mechanism is probably much less complex because the entire pB15 genome (50 kb) is smaller than the R64 shufflon (D. GUTTMAN and P. TURNER, unpublished data). However, pB15 might also feature two (or more) types of sex pili, and the influence of cell density and/or glucose concentration on the expression of genes related to pilus formation in pB15 is unknown. For instance, these factors might cause pB15 to differentially express one or more of its pilus genes, thereby influencing invasiveness. Because this mechanistic explanation is highly speculative, I discuss several other possibilities below. In addition, I relate my data to previous studies in plasmid biology.

Potential mechanisms for density-dependent conjugation:
A simple explanation for the density-dependent effect is that the conjugation process in pB15 somehow becomes "saturated" at higher cell densities in static culture, much as bacterial growth reaches a maximum rate that cannot be raised by increasing the concentration of a limiting resource (MONOD 1949 ). However, saturation kinetics cannot explain the apparent maximum rate of plasmid increase at intermediate densities, as observed here. A similar possibility is that conjugation rate in pB15 is not a function of cell density per se but instead responds to the concentration of glucose, which was varied to manipulate cell density. That is, higher concentrations of glucose may somehow inhibit the conjugal transfer of pB15.

I examined this possibility by looking closely at the population dynamics occurring in the 24-hr mating experiments used to estimate conjugation rate. There was some evidence that the number of transconjugants increases unexpectedly during the transition from exponential growth to stationary phase, irrespective of cell density, suggesting that conjugation in pB15 is somehow stimulated by the depletion of glucose (data not shown). In particular, the number of transconjugants measurably increased between 8 and 10 hr of the growth cycle, whereas by this time the donors and recipients had evidently made the transition from exponential growth to stationary phase as the medium was being depleted of glucose. Simple models assume that the rates of bacterial growth and plasmid transfer are Monod functions (MONOD 1949 ) of resource concentration, so that when resources are exhausted growth and conjugation cease (STEWART and LEVIN 1977 ; SIMONSEN et al. 1990 ). Thus, the conjugation rate of a plasmid is expected to be proportional to the growth rate of the mating population. However, in IncP plasmids the expression of transfer proteins can be highest when host cells are growing slowly and thus when vertical transfer of plasmids is minimized (PANSEGRAU et al. 1994 ). Further experiments could examine the dynamics of 24-hr matings with pB15 and the possibility of unexpectedly high transfer of the plasmid during the transition to stationary phase in static culture.

High cell densities (or high glucose concentrations) might strongly impact other plasmid or bacterial traits in my experiments. For instance, the average number of pB15 copies per cell might somehow decline at high densities in static environments, due to changes in the host (e.g., replication control during cell division) or in plasmid regulation of copy number. But this explanation works only if changes in plasmid copy number affect infectiousness (e.g., if increased copy number increases conjugative pilus formation), and to my knowledge this link has not been documented. Similarly, the level of carbon source might negatively impact the motility of cells in my static-culture experiments, thus reducing the potential for transfer to occur at high cell densities. This idea could explain the pB15 results, but seems unlikely given that transfer of R1-drd19 was not similarly hampered by elevated densities in static culture.

Relevance to previous work:
Most models of plasmid transfer depend on simple mass-action kinetics, but it is now widely recognized that the majority of bacteria found in natural, clinical, and industrial settings persist in association with surfaces (DAVEY and O'TOOLE 2000 ). Therefore, plasmid conjugation is likely to be a complex process that extends beyond simple models (e.g., LAGIDO et al. 2003 ). Further complexities may arise because conjugation can be sensitive to both abiotic and biotic conditions (FERNANDEZ-ASTORGA et al. 1992 ), such as temperature (FERNANDEZ-TRESGUERRES et al. 1995 ) and dissimilarities between donor and recipient strains (e.g., DIONISIO et al. 2002 ). PAPPAS and WINANS 2003 showed that conjugation of the Ti plasmid is activated by quorum sensing in the phyto-pathogen Agro-bacterium tumefaciens, demonstrating density-dependent changes in plasmid transfer ability. Thus, earlier findings indicate that phenotypic plasticity can occur in plasmids, but to my knowledge the data were never described in light of this important biological phenomenon.

Virulence models predict that virulent parasites can evolve at increased host densities, because available host abundance favors selection for highly infectious genotypes. The underlying assumption is that increased virulence leads to reduced host fitness, thereby preventing a parasite from simultaneously maximizing vertical and horizontal transfer. We chose plasmid pB15 to examine this hypothesis (TURNER et al. 1998 ), on the basis of its ability to successfully invade E. coli K-12 populations growing at 10⁸ cells/ml in chemostat culture (LUNDQUIST and LEVIN 1986 ). To do so, we used static serial-culture environments containing 1000 µg/ml glu (10⁹ cells/ml). Our previous experiment was correctly designed to test the theoretical prediction, in principle. Preliminary experiments with pB15 showed that the plasmid can invade when rare at moderate cell densities (10⁸ cells/ml; 50 µg/ml glu) in static culture (TURNER 1995 ), as confirmed here. However, my current data also reveal that the combination of plasmid pB15 and a glucose-rich static culture regime was an unfortunate choice, because this environment negatively impacts the plasmid's conjugation ability (i.e., R1-drd19 would have been an okay choice because its spread improves with increasing cell density in the presence and absence of shaking). Hence, we may have created weak ability for host density to select for conjugation in our prior study, resulting in the apparently stochastic evolution of increased (and reduced) conjugation in all of our treatments. We can be criticized for conducting our study at 1000 µg/ml glu, without first examining the associated invasion dynamics of pB15 (TURNER et al. 1998 ); otherwise we might have eliminated pB15 or static culture as suitable choices. But the negative effect of cell density on conjugation rate of pB15 in static culture is both unusual and unexpected. Therefore, we had no a priori reason to believe that a 20-fold increase in cell density would inhibit the invasiveness of pB15 seen in our preliminary assays (TURNER 1995 ).

Whatever the precise mechanistic explanation for the observed phenomena, they may explain the failure of a simple model to predict the evolutionary response of pB15 to experimental manipulations of susceptible host density. Although the genetic assumption of a trade-off between rates of horizontal and vertical transmission was fulfilled (TURNER et al. 1998 ), the ecological assumption that the rate of horizontal transmission is simply proportional to susceptible host density was evidently not satisfied (current findings). More generally, models of phenotypic evolution depend on both genetic and ecological assumptions, and the predictions of these models may fail as a consequence of violating either type of assumption.

	ACKNOWLEDGMENTS

I thank M. Girgenti and R. Montville for technical assistance and the members of my laboratory group for enlightening discussion. R. Lenski, B. Levin, and J. Smith kindly provided strains. T. Cooper, V. Cooper, S. Duffy, R. Lenski, J. Wertz, and two anonymous reviewers gave helpful advice and valuable comments on the manuscript. This work was supported by grants from the National Science Foundation (DEB-02-01860 and DEB-01-29089).

Manuscript received December 5, 2003; Accepted for publication January 27, 2004.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

BALE, M. J., J. C. FRY, and M. J. DAY, 1987  Plasmid transfer between strains of Pseudomonas aeruginosa on membrane filters attached to river stones. J. Gen. Microbiol. 133:3099-3107.[Medline]

BERGSTROM, C. T., M. LIPSITCH, and B. R. LEVIN, 2000  Natural selection, infectious transfer and the existence conditions for bacterial plasmids. Genetics 155:1505-1519.[Abstract/Free Full Text]

BJORKLOF, K., A. SUONIEMI, K. HAAHTELA, and M. ROMANTSCHUK, 1995  High-frequency of conjugation versus plasmid segregation of RP1 in epiphytic Pseudomonas syringae populations. Microbiology 141:2719-2727.[Abstract]

BOUMA, J. E. and R. E. LENSKI, 1988  Evolution of a bacteria/plasmid association. Nature 335:351-352.[CrossRef][Medline]

BRADLEY, D. E., 1983  Derepressed plasmids of incompatibility group I1 determine two different morphological forms of pilus. Plasmid 9:331-334.[CrossRef][Medline]

BRADLEY, D. E., 1984  Characteristics and function of thick and thin conjugative pili determined by transfer-derepressed plasmids of incompatibility groups I₁, I₂, I₅, B, K, and Z. J. Gen. Microbiol. 130:1489-1502.[Medline]

BRADSHAW, A. D., 1965  Evolutionary significance of phenotypic plasticity in plants. Adv. Genet. 13:115-155.

BULL, J. J., 1994  Perspective: virulence. Evolution 48:1423-1437.[CrossRef]

BULL, J. J., I. J. MOLINEUX, and W. R. RICE, 1991  Selection of benevolence in a host-parasite system. Evolution 45:875-882.[CrossRef]

CARLTON, B. C., and B. J. BROWN, 1981 Gene mutation, pp. 222–242 in Manual of Methods for General Bacteriology, edited by P. GERHARDT. American Society for Microbiology, Washington, DC.

CLEWELL, D. B., 1993 Bacterial Conjugation. Plenum Press, New York.

DAVEY, M. E. and G. A. O'TOOLE, 2000  Microbial biofilms: from ecology to molecular genetics. Microbiol. Mol. Biol. Rev. 64:847-867.[Abstract/Free Full Text]

DEARSLEY, A. L., R. E. SINDEN, and I. A. SELF, 1990  Sexual development in malarial parasites, gametocyte production, fertility and infectivity to the mosquito vector. Parasitology 100:359-368.

DEVANAS, M. A., D. RAFAELIESHKOL, and G. STOTZKY, 1986  Survival of plasmid-containing strains of Escherichia coli in soil—effect of plasmid size and nutrients on survival of hosts and maintenance of plasmids. Curr. Microbiol. 13:269-277.[CrossRef]

DIFFLEY, P., J. O. SCOTT, K. MAMA, and T. N.-R. TSEN, 1987  The rate of proliferation among African trypanosomes is a stable trait that is directly related to virulence. J. Trop. Med. Hyg. 36:533-540.

DIONISIO, F., I. MATIC, M. RADMAN, O. RODRIGUES, and F. TADDEI, 2002  Plasmids spread very fast in heterogeneous bacterial communities. Genetics 162:1525-1532.[Abstract/Free Full Text]

EBERT, D. and J. J. BULL, 2003  Challenging the trade-off model for the evolution of virulence: Is virulence management feasible? Trends Microbiol. 11:15-20.[CrossRef][Medline]

EBERT, D. and K. L. MANGIN, 1997  The influence of host demography on the evolution of virulence of a microsporidian gut parasite. Evolution 51:1828-1837.[CrossRef]

EWALD, P. W., 1994 Evolution of Infectious Disease. Oxford University Press, New York.

FERNANDEZ-ASTORGA, A., A. MUELA, R. CISTERNA, J. IRIBERRI, and I. BARCINA, 1992  Biotic and abiotic factors affecting plasmid transfer in Escherichia coli strains. Appl. Environ. Microbiol. 58:392-398.[Abstract/Free Full Text]

FERNANDEZ-TRESGUERRES, M. E., M. MARTIN, D. GARCIA DE VIEDMA, R. GIRALDO, and R. DIAZ-OREJAS, 1995  Host growth temperature and a conservative amino acid substitution in the replication protein of pPS10 influence plasmid host range. J. Bacteriol. 177:4377-4384.[Abstract/Free Full Text]

GANUSOV, V. V. and A. V. BRILKOV, 2002  Estimating the instability parameters of plasmid-bearing cells. I. Chemostat culture. J. Theor. Biol. 219:193-205.[CrossRef][Medline]

GERDES, K., P. B. RASMUSSEN, and S. MOLIN, 1986  Unique type of plasmid maintenance: postsegregational killing of plasmid-free cells. Proc. Natl. Acad. Sci. USA 83:3116-3120.[Abstract/Free Full Text]

HEDGES, R. W. and N. DATTA, 1973  Plasmids determining I pili constitute a compatibility group. J. Gen. Microbiol. 77:19-25.[Medline]

HELINSKI, D. R., A. E. TOUKDARIAN and R. P. NOVICK, 1996 Replication control and other stable maintenance mechanisms of plasmids, pp. 2295–2324 in Escherichia coli and Salmonella: Cellular and Molecular Biology, Ed. 2, edited by F. C. NEIDHARDT. American Society for Microbiology, Washington, DC.

HERRE, E. A., 1993  Population structure and the evolution of virulence in nematode parasites of fig wasps. Science 259:1442-1445.[Abstract/Free Full Text]

KOMANO, T., 1999  Shufflons: multiple inversion systems and integrons. Annu. Rev. Genet. 33:171-191.[CrossRef][Medline]

KOMANO, T., S.-R. KIM, and T. YOSHIDA, 1995  Mating variation by DNA inversions of shufflon in plasmid R64. Adv. Biophys. 31:181-193.[CrossRef][Medline]

LAGIDO, C., I. J. WILSON, L. A. GLOVER, and J. I. PROSSER, 2003  A model for bacterial conjugal transfer on solid surfaces. FEMS Microbiol. Ecol. 44:67-78.[CrossRef]

LEDERBERG, J. and E. L. TATUM, 1946  Gene recombination in Escherichia coli. Nature 158:558.

LENSKI, R. E and J. E. BOUMA, 1987  Effects of segregation and selection on instability of plasmid pACYC184 in Escherichia coli B. J. Bacteriol. 169:5314-5316.[Abstract/Free Full Text]

LENSKI, R. E., M. R. ROSE, S. C. SIMPSON, and S. C. TADLER, 1991  Long-term experimental evolution in Escherichia coli. I. Adaptation and divergence during 2,000 generations. Am. Nat. 138:1315-1341.[CrossRef]

LEVIN, B. R., 1980 Conditions for the existence of R-plasmids in bacterial populations, pp. 197–202 in Antibiotic Resistance: Transposition and Other Mechanisms, edited by S. MITSUHASHI, L. ROSIVAL and V. KRCMERY. Springer-Verlag, Berlin.

LEVIN, B. R., F. M. STEWART, and L. CHAO, 1977  Resource-limited growth, competition, and predation: a model and experimental studies with bacteria and bacteriophage. Am. Nat. 111:3-24.[CrossRef]

LEVIN, B. R., F. M. STEWART, and V. A. RICE, 1979  The kinetics of conjugative plasmid transmission: fit of a simple mass action model. Plasmid 2:247-260.[CrossRef][Medline]

LICHT, T. R., B. B. CHRISTENSEN, K. A. KROGFELT, and S. MOLIN, 1999  Plasmid transfer in the animal intestine and other dynamic bacterial populations: the role of community structure and environment. Microbiology 145:2615-2622.[Abstract/Free Full Text]

LUNDQUIST, P. E. and B. R. LEVIN, 1986  Transitory derepression and the maintenance of conjugative plasmids. Genetics 113:483-497.[Abstract/Free Full Text]

MAY, R. M. and R. M. ANDERSON, 1983  Epidemiology and genetics in the coevolution of parasites and hosts. Proc. R. Soc. Lond. Ser. B 219:281-313.[Medline]

MESSENGER, S. L., I. J. MOLINEUX, and J. J. BULL, 1999  Virulence evolution in a virus obeys a trade-off. Proc. R. Soc. Lond. Ser. B 266:397-404.[Medline]

MOENNE-LOCCOZ, Y. and R. W. WEAVER, 1995  Plasmids and saprophytic growth of Rhizobium-leguminosarum bv. trifolii W14-2 in soil. FEMS Microbiol. Evol. 18:139-144.[CrossRef]

MONOD, J., 1949  The growth of bacterial cultures. Annu. Rev. Microbiol. 3:371-394.[CrossRef]

NGUYEN, T. N. M., Q. G. PHAN, L. P. DUONG, K. P. BERTRAND, and R. E. LENSKI, 1989  Effects of carriage and expression of the Tn10 tetracycline-resistance operon on the fitness of Escherichia coli K12. Mol. Biol. Evol. 6:213-225.[Abstract]

NORDSTROM, K., S. MOLIN, and H. AAGAARD-HANSEN, 1980  Partitioning of plasmid R1 in Escherichia coli. I. Kinetics of loss of plasmid derivatives deleted of the par region. Plasmid 4:215-227.[CrossRef][Medline]

PANSEGRAU, W., E. LANKA, P. T. BARTH, D. H. FIGURSKI, and D. G. GUINEY et al., 1994  Complete nucleotide sequence of Birmingham IncP plasmids: compilation and comparative analysis. J. Mol. Biol. 239:623-663.[CrossRef][Medline]

PAPPAS, K. M. and S. C. WINANS, 2003  A LuxR-type regulator from Agrobacterium tumefaciens elevates Ti plasmid copy number by activating transcription of plasmid replication genes. Mol. Microbiol. 48:1059-1073.[CrossRef][Medline]

PAULSSON, J., 2002  Multileveled selection on plasmid replication. Genetics 161:1373-1384.[Abstract/Free Full