Growth-Dependent Predation and Generalized Transduction of Antimicrobial Resistance by Bacteriophage

Abstract

Bacteriophage (phage) are both predators and evolutionary drivers for bacteria, notably contributing to the spread of antimicrobial resistance (AMR) genes by generalized transduction. Our current understanding of this complex relationship is limited. We used an interdisciplinary approach to quantify how these interacting dynamics can lead to the evolution of multidrug-resistant bacteria. We cocultured two strains of methicillin-resistant Staphylococcus aureus, each harboring a different antibiotic resistance gene, with generalized transducing phage. After a growth phase of 8 h, bacteria and phage surprisingly coexisted at a stable equilibrium in our culture, the level of which was dependent on the starting concentration of phage. We detected double-resistant bacteria as early as 7 h, indicating that transduction of AMR genes had occurred. We developed multiple mathematical models of the bacteria and phage relationship and found that phage-bacteria dynamics were best captured by a model in which phage burst size decreases as the bacteria population reaches stationary phase and where phage predation is frequency-dependent. We estimated that one in every 10⁸ new phage generated was a transducing phage carrying an AMR gene and that double-resistant bacteria were always predominantly generated by transduction rather than by growth. Our results suggest a shift in how we understand and model phage-bacteria dynamics. Although rates of generalized transduction could be interpreted as too rare to be significant, they are sufficient in our system to consistently lead to the evolution of multidrug-resistant bacteria. Currently, the potential of phage to contribute to the growing burden of AMR is likely underestimated. IMPORTANCE Bacteriophage (phage), viruses that can infect and kill bacteria, are being investigated through phage therapy as a potential solution to the threat of antimicrobial resistance (AMR). In reality, however, phage are also natural drivers of bacterial evolution by transduction when they accidentally carry nonphage DNA between bacteria. Using laboratory work and mathematical models, we show that transduction leads to evolution of multidrug-resistant bacteria in less than 8 h and that phage production decreases when bacterial growth decreases, allowing bacteria and phage to coexist at stable equilibria. The joint dynamics of phage predation and transduction lead to complex interactions with bacteria, which must be clarified to prevent phage from contributing to the spread of AMR.

Keywords: Staphylococcus aureus; antimicrobial resistance; bacteriophages; horizontal gene transfer; mathematical modelling; microbiology; transduction.

FIG 1

Phage lytic cycle and generalized transduction. In this environment, only some bacteria carry an antimicrobial resistance (AMR) gene (shown in green). The lytic cycle starts when a lytic phage infects a bacterium by binding and injecting its DNA (1). Phage molecules degrade bacterial DNA and utilize bacterial resources to create new phage components and replicate (2). These components are then assembled to form new phage particles (3). At this stage, bacterial DNA left in the cell can be packaged by mistake instead of phage DNA, which creates a transducing phage and starts the process of generalized transduction. In our example, the transducing phage carries the AMR gene. After a latent period of typically several minutes, the phage trigger lysis of the bacterium, bursting it and releasing the phage (4). The transducing phage can infect another bacterium, binding and injecting the AMR gene it is carrying (5). If this gene is successfully integrated into the bacterial chromosome (6), this creates a new transductant bacterium carrying this AMR gene (7). Note that the transduced bacterial DNA could also be a plasmid, in which case it would circularize instead of integrating into the chromosome of the transductant bacterium. The figure is not to scale.

FIG 2

Starting concentration of exogenous phage 80α affected the equilibrium values of phage and bacteria in our cocultures. The starting concentration of both single-resistant S. aureus parent strains (B_E for erythromycin and B_T for tetracycline) was 10⁴ CFU per mL. Each panel shows the results with a different starting concentration of exogenous phage (P_L): either 10³, 10⁴, or 10⁵ plaque-forming units (PFU) per mL. We detected double-resistant progeny (B_ET) as early as 7 h, indicating that transduction occurred rapidly. Error bars indicate means ± standard errors from 3 experimental replicates. There are no data for the time period of 9 h to 15 h.

FIG 3

80α lysogeny does not occur at a detectable level in our coculture. (a) Cocultures with bacteria not exposed or previously exposed to phage. The starting concentration of both single-resistant S. aureus parent strains (B_E for erythromycin and B_T for tetracycline) was 10⁴ CFU per mL, and the starting concentration of exogenous phage 80α (P_L) was 10⁴ PFU per mL. double-resistant progeny (B_ET) are generated by transduction. The initial coculture was diluted in fresh media after 24 h to form a new coculture with bacteria previously exposed to phage. Phage were added in the new coculture to reach a concentration of 10⁴ PFU/mL. Error bars indicate means ± standard errors, from 3 experimental replicates. (b) Confirmation of absence of detectable lysogeny by polymerase chain reaction. DNA was extracted from the cocultures after 24 h. S. aureus RN4220 strains lysogenic and nonlysogenic for 80α were used as positive and negative controls. L, ladder; attL, left prophage junction; attR, right prophage junction; attB, bacterial insertion site. Detection of attL and attR indicates that prophage are present in the DNA, while detection of attB indicates the presence of bacteria not lysogenic for 80α.

FIG 4

Phage predation and generalized transduction model diagram and different phage-bacteria interactions considered. (a) Model diagram. Each bacteria strain (B_E, resistant to erythromycin; B_T, resistant to tetracycline; B_ET, resistant to both) can replicate (purple). The lytic phage (P_L) multiply by infecting a bacterium and bursting it to release new phage (gold). This process can create transducing phage (P_E or P_T) carrying a resistance gene [erm(B) or tet(K), respectively] taken from the infected bacterium (green). These transducing phage can then generate new DRP (B_ET) by infecting the bacterial strain carrying the other resistance gene (green). (b) Phage predation in the model is either density- or frequency-dependent. (Top) With a density-dependent interaction, the number of infections scales linearly with the number of phage and bacteria. (Bottom) A frequency-dependent interaction illustrates that some phage may not infect a bacterium or that multiple phage may infect the same bacterium. (c) Phage predation in the model can decrease as bacterial growth decreases. (Top) A change in bacterial growth phase can affect surface receptors, leading to a reduced phage adsorption rate. (Bottom) Since phage replication relies on bacterial processes, reduced bacterial growth can translate into reduced phage burst size. (d) Proposed function linking phage predation parameters to bacterial growth. This shows the multiplier applied to decrease phage parameters as the bacterial population increases toward carrying capacity, equivalent to a decrease in bacterial growth. Here, the carrying capacity is 2.76 × 10⁹ CFU/mL, estimated from our data.

FIG 5

Accuracy of the best-fitted models to reproduce in vitro phage-bacteria dynamics. (a and b) The models with only phage burst size linked to bacterial growth are the most accurate to reproduce in vitro trends in lytic phage (a) and double-resistant bacteria (b) numbers, starting from a bacterial concentration of 10⁴ CFU/mL and varying phage concentrations. All models (dashed lines) can reproduce the trends seen in vitro when phage are started at 10³ or 10⁴ PFU/mL (data in solid lines), but only the models with just the phage burst size linked to bacterial growth (colored model output) can reproduce the trend seen when phage are started at 10⁵ PFU/mL. Other models (gray) have only the phage adsorption rate linked to bacterial growth or both the phage adsorption rate and burst size. Models are fitted to the 10³ and 10⁵ data and tested with the 10⁴ data. Parameter values used are the median fitted values (Table 1). Shaded areas indicate standard deviations generated from Poisson resampling of model results. Error bars for the data (solid lines) indicate means ± standard errors from 3 experimental replicates. (c) When further testing fitted model dynamics starting from 10⁶ CFU/mL bacteria and varying phage concentrations, the density-dependent model incorrectly predicts bacterial extinction, while the frequency-dependent model reproduces the trend but not the exact values of the 24 h data. In the coculture used to generate the data, each single-resistant parent strain (B_E and B_T) is added at a starting concentration of 10⁶ CFU/mL, and no DRP (B_ET) are initially present. The starting concentration of lytic phage (P_L) varies (x axis). Points indicate mean results and are each slightly shifted horizontally to facilitate viewing. Error bars indicate either means ± standard deviation for the models (left/center) or means ± standard errors for the data (right). Parameter values used are the median fitted values (Table 1).

FIG 6

Underlying phage and bacteria dynamics generated by the best-fitting frequency-dependent model with burst size linked to bacterial growth. Model parameters are the median estimates from model fitting (Table 1). (a) Phage burst size over time by starting phage concentration. As bacteria reach stationary phase after 8 h, phage burst size decreases. In the 10⁵ data set, we see that burst size is predicted to increase again after 20 h. This is due to bacterial numbers decreasing as bacteria are being lysed by phage. (b) Relative change in phage and bacterial numbers over time by starting phage concentration. The number of new phage generated at each time step increases (positive value) until bacteria reach stationary phase around 8 h. This applies to lytic and transducing phage. In the 10⁵ data set, phage keep increasing after 10 h, eventually causing a decrease in bacterial numbers (negative value), which translates into a further acceleration in the increase in phage numbers due to the increased burst size (Fig. 5a). After 8 h, the relative changes in lytic and transducing phage numbers are identical. (c) Incidence of lytic (gold) and transducing (green) phage over time by starting phage concentration (line type). For any data set and time point, there is approximately 1 new transducing phage generated for each 10⁸ new lytic phage. (d) Fraction of DRP generated by transduction each hour over time by starting phage concentration (line type). DRP generation always occurs predominantly by transduction rather than by growth of already existing DRP. Note that the time at which DRP are first generated varies by starting phage concentration.