Conjugation dynamics depend on both the plasmid acquisition cost and the fitness cost

Abstract

Plasmid conjugation is a major mechanism responsible for the spread of antibiotic resistance. Plasmid fitness costs are known to impact long-term growth dynamics of microbial populations by providing plasmid-carrying cells a relative (dis)advantage compared to plasmid-free counterparts. Separately, plasmid acquisition introduces an immediate, but transient, metabolic perturbation. However, the impact of these short-term effects on subsequent growth dynamics has not previously been established. Here, we observed that de novo transconjugants grew significantly slower and/or with overall prolonged lag times, compared to lineages that had been replicating for several generations, indicating the presence of a plasmid acquisition cost. These effects were general to diverse incompatibility groups, well-characterized and clinically captured plasmids, Gram-negative recipient strains and species, and experimental conditions. Modeling revealed that both fitness and acquisition costs modulate overall conjugation dynamics, validated with previously published data. These results suggest that the hours immediately following conjugation may play a critical role in both short- and long-term plasmid prevalence. This time frame is particularly relevant to microbiomes with high plasmid/strain diversity considered to be hot spots for conjugation.

Keywords: antibiotic resistance; conjugation; fitness cost; horizontal gene transfer; mathematical modeling.

Figure 1. Transconjugants exhibit an acquisition cost following conjugation

1. Schematic of conjugation whereby plasmid DNA from a donor (D, blue) is transferred to a recipient (R, red), generating a transconjugant (T, yellow). R and D are each resistant to an antibiotic (A₁ and A₂, respectively), but sensitive to the other. The transconjugant (T) is uniquely resistant to both antibiotics.

2. Conjugation protocol involves mixing D and R, followed by one‐hour incubation at 25°C. Cells are then diluted into media containing A₁ and A₂ and growth is captured over time in a 96‐well plate.

3. OD₆₀₀ for de novo T (aqua) and adapted T (gray), (RP4 transconjugants) initiated with the same number of cells per well. Each curve is a biological replicate. Black dashed lines are best‐fits.

4. Growth rate (left) and lag time (right) for the plasmid‐free recipient R, adapted T, de novo T after 1 h of conjugation, and de novo T diluted and re‐grown after 24 h. De novo T growth rates (aqua) are statistically less than adapted T (gray) (P = 1.12e‐08, Appendix Table S2) and plasmid‐free cells (red) (P = 7.27e‐09, Appendix Table S2). De novo T lag times are statistically greater than adapted T (P = 5.71e‐08 Appendix Table S2) and plasmid‐free cells (red) (3.77e‐09, Appendix Table S2). Data represent biological replicates. All statistics were done using a one‐way ANOVA with Bonferroni correction.

5. Left: De novo T growth rates (aqua) are statistically less than adapted T growth rates (gray) (P = 7.34e‐05 and 4.95e‐05 for 15 and 120 min, respectively). Right: Lag times were multiplied by true T₀ and divided by the mean adapted T lag (Appendix Fig S1D is non‐normalized). De novo T (aqua) normalized lag times are statistically less than adapted T (gray) (P = 1.30e‐09 and 1.31e‐09 for 15 and 120 min, respectively). Data represent biological replicates. All statistics were done using a one‐way ANOVA with Bonferroni correction.

6. After 24 h, each condition (E) was diluted and re‐grown. Left: growth rates are statistically identical (P = 1.00 for both 15 and 120 min). Right: All lag times are statistically identical (P = 0.69 and P = 0.48 for 15 and 120 min, respectively). Lag times normalized as described in (E). All data represent biological replicates. All statistics were done using a one‐way ANOVA with Bonferroni correction.

Source data are available online for this figure.

Figure 2. Acquisition cost quantification for RP4

1. The time (t^*, orange) at which a specified cell density threshold (T^*, top purple) is reached uniquely depends on the initial cell density (T₀, bottom purple), and the growth rate (µ, aqua) and lag times (λ, orange). Assuming background subtraction, the line can be represented by the equation that is shown.

2. Representative standard curve generation is shown. Left: To generate the standard curve, adapted T are diluted in 10‐fold increments and growth is measured over time (dark to light gray is T₀). Circles indicate the t^* (purple line) corresponding to T^* (orange line). Aqua line represents the out‐growth of T following a conjugation experiment (i.e., de novo T). Right: The initial cell density is plotting against t ^*; black line indicates the standard curve.

3. Left: RP4 adapted T growth initiated with decreasing true T₀ (dark to light gray); blue markers show the time to reach OD₆₀₀ of 0.275 (t^*). Right: Standard curve is shown in blue. Error bars are standard deviation of three biological replicates.

4. True and predicted CFU of RP4 with the recipient E. coli strain MG1655. Scatter points represent three biological replicates, and bar height is the average.

5. True and predicted CFU of RP4 with the recipient K. pneumoniae (KPN) strain AL2425. Scatter points represent four biological replicates, and bar height is the average.

Source data are available online for this figure.

Figure 3. Generality of acquisition cost

1. OD₆₀₀ for de novo (aqua) and adapted transconjugants (gray) for the plasmid pR are shown over time. Black lines are best‐fits. Individual curves are biological replicates.

2. True and predicted CFU for the plasmid pR are statistically identical (P = 0.34 and 0.86 for one and two‐tailed t‐tests, respectively). Scatter points represent biological replicates, and bar height is the average.

3. Growth rates are shown for adapted T carrying RP4 under variable glucose (glu) and casamino acid (caa) concentrations. Values represent % w/v. Scatter points represent biological replicates.

4. Acquisition costs were quantified for the same glucose and casamino acid concentrations from (C). Scatter points represent the average, and error bars represent standard deviation, of three biological replicates. Aqua and red indicate glucose at 0.4% and 0.04% w/v, respectively. Y‐axis is acquisition cost (T _pred/T ₀) normalized to the cost in the absence of casamino acids.

5. T _pred compared to T₀ for six well‐characterized plasmids. Representatives are shown of two biological replicates (see Appendix Fig S10 for day‐to‐day variability). R1, R1drd, and pRK100 do not show a significant acquisition cost (P = 0.93, 0.79, and 0.28, respectively), whereas RIP113, R6K, and R6Kdrd do (P = 6.79e‐05, 7.57e‐05, and 0.037, respectively, one‐tailed t‐test, Appendix Table S3A).

6. T _pred was significantly less than T₀ for clinical plasmids p41, p168, p193, and p283 at 37°C (P = 7.10e‐05, 2.10e‐05, 0.021, 1.90e‐05, respectively, n = 4, 4, 3, 2, respectively, one‐tailed t‐test, Appendix Table S3A). In all cases except p283, bars represent averages and scatter points individual measurements from at least three biological replicates; p283 has two biological replicates.

7. T _pred for two clinical plasmids, p193 and p168, at 30°C was significantly less than T₀ (P = 2.80e‐04, 2.72e‐04, respectively, n = 3, 4, respectively, one‐tailed t‐test, Appendix Table S3A).

8. All acquisition costs scattered against the fitness cost measured under the identical condition for each plasmid. Acquisition costs are measured as the ratio between T _pred/T ₀. Black line is the linear regression line of best fit, and R ² = 0.01 (shown in the bottom left). Error bars represent standard deviation; the type of replicates used for these error bars is listed in Appendix Table S3A.

Source data are available online for this figure.

Figure 4. Conjugation model incorporating acquisition cost

1. Network diagram of conjugation model. The plasmid‐free population (S⁰) acquires the plasmid from the plasmid‐adapted population (S^A), turning into a transient de novo transconjugant (S^D) at a rate η (the conjugation efficiency). Finally, S^A can revert to S⁰ according to the plasmid segregation error rate κ. The de novo population, in turn, transitions into the adapted population at the rate β. Growth rates for S⁰ and S^D are scaled relative to the plasmid‐adapted population (µ) based on the scalars α and ρ, respectively. Not included in diagram: dilution of all populations out of the system at rate D.

2. RP4 data were fit to the model to calculate β and ρ. Dotted line shows model fit. Data are from Fig 1C.

3. Model predicts accurate growth rates and lag times based on fitted parameters.

4. Parameter sensitivity to ρ. Bottom: Average observed growth rate is defined as (S^Dρµ+ S^Aµ)/(S^D + S^A), and is measured for increasing β from light to dark. Top: Corresponding population density of S¹ over time. β is fixed to 0.01 based on RP4 fitting.

5. Parameter sensitivity to β. Bottom: Average observed growth rate measured for increasing β from light to dark where ρ = 0.3 (left) or ρ = 0 (right). Top: Corresponding population density of S¹ over time.

6. RP4 data (Fig 4B) are re‐fit with fixed ρ (top) and fixed β (bottom).

Figure 5. Fitness cost versus acquisition cost

1. Long‐term temporal dynamics of S^A, S⁰, and S^D are shown in red, blue, and gray, respectively, from the main model (Fig 4A). X‐axis is time over 21 days, and y‐axis is the fraction of each population.

2. A population is initiated with a 1:1 ratio of S⁰ and S^A and the total plasmid‐carrying population fraction (S¹ = S^A+S^D) is tracked over time. Left: β is held constant (β = 0.01) and α is increased from no cost (α = 1) to high cost (α = 1.5). Right: α is held constant (α = 1.2) and β is increased from slow transition (β = 10⁻⁴) to rapid transition (β = 1).

3. Heat map shows where the observed growth rate of S¹ (μ_obs, calculated using Appendix equation S6) differed from the maximum growth rate under ideal conditions (e.g., μ, if there is no acquisition cost). A 98% threshold was used to numerically define the region where μ_obs differed significantly from μ (e.g., μ_obs/μ < 0.98). Any α and β combination meeting this criterion is colored blue and are red otherwise. Changing this threshold did not qualitatively change conclusions (Appendix Fig S9). Changing the conjugation efficiency (η) shifts the boundary (increasing from light to dark shades of blue).

4. Validation of modeling predictions using four plasmids from left to right: RP4 (in this study), p193, p41, and p168 (from previous work). Data are reproduced with permission from Nature Communications (Lopatkin et al, 2017), under the Creative Commons Attribution 4.0 International License. Marker shapes and colors were modified for visualization purposes. Solid line shows original model fit (e.g., Appendix equation S1‐S2). Dotted lines show updated model fit (e.g., Appendix equations S3‐S5). Experiments were performed at least twice. Error bars represent the standard deviation of four to six measurements.