Strategies for antibiotic  administration typically follow the “hit hard and hit quick” paradigm, where the objective is to reduce pathogen loads as rapidly as possible. While the rationale for this is clear, we posited that the “hit hard and hit quick” paradigm might have negative consequences on the spread of antibiotic resistance into new microbial genotypes. Whenever antibiotics are applied, the genotypes carrying antibiotic resistance genes become predominant and the likelihood that antibiotic resistance genes spread increases. Our rationale was based on two main observations: 1) the spatial arrangement of microbial genotypes changes over time, and 2) antibiotic resistance genes can often be transferred between neighboring genotypes. So what does a temporal view on microbial spatial arrangement have to do with antibiotic resistance spread?

Consider that most microbial infections of clinical importance are surface-associated (e.g. biofilms) and that antibiotic resistance is often encoded on metabolically costly plasmids. These plasmids can transfer between different genotypes in a cell contact-dependent process termed conjugation. In the absence of antibiotics, new genotypes that receive the plasmid suffer a fitness cost relative to their plasmid-free versions, so they become increasingly rare over time. This means that the load of antibiotic resistance genes that can be transferred decreases as the community expands. This leads to the expectation that delaying antibiotic administration will lower the chances that these resistance genes can spread. Moreover, as a collection of microbial genotypes grows and expands across space, the different genotypes become increasingly spatially segregated from each other. Because spatial segregation reduces the number of cell-cell contacts between different genotypes, it will also reduce the probability of plasmid transfer between those genotypes. Over time, the reduction in cell-cell contacts will also lower the chances for antibiotic resistance genes to spread throughout the community. These considerations made us think that “hit hard but perhaps wait a while” could be an alternative approach for those infections with large numbers of antibiotic resistant genotypes. While adhering to the “hit hard and hit quick” paradigm is undoubtedly important for reducing pathogen loads, it could also maximize the ability of plasmids to proliferate within new genotypes and contribute to the spread of antibiotic resistance.

 We tested this hypothesis using controlled laboratory experiments consisting of a plasmid donor and a potential recipient strain of Pseudomonas. To our surprise, we did not observe a negative relationship between the timing of antibiotic administration and the proliferation of plasmids within new genotypes. Instead, we observed a unimodal relationship where plasmid proliferation is maximized at intermediate antibiotic administration times (Fig. 1). Why is this? What causes a non-trivial unimodal relationship to emerge? We postulated that this is caused by the dynamic interplay between the processes of plasmid loss and gain. At early antibiotic administration times, there is likely insufficient time for plasmid transfer to occur, resulting in the absence of new genotypes containing the plasmid. At late antibiotic administration times, the plasmid has been purged from the system, again resulting in the absence of new genotypes containing the plasmid. It is only at intermediate antibiotic administration times, when there is sufficient time for plasmid transfer to occur but not enough time for the plasmid to be purged from the system, that we observe significant transfer of the plasmid into new genotypes.

Fig. 1:  Effect of the timing of chloramphenicol administration after the onset of range expansion on plasmid spread. Representative microscopy images of spatial patterns formed after chloramphenicol was administered at different times after the onset of range expansion. Plasmid donors, magenta; potential recipients, green; transconjugants, cyan; donors cured of the plasmid, red.

To expand on this hypothesis, we performed individual-based computational modelling to quantitatively describe how combinations of plasmid transfer and loss probabilities affect the spread of a plasmid into new genotypes (Fig. 2). To our surprise, we found a wide range of possible relationships between the timing of antibiotic administration and the spread of antibiotic resistance-encoding plasmids. When plasmid loss is much higher than plasmid transfer, we observed a monotonically decreasing relationship as we initially expected. This represents a scenario where the plasmid is lost more rapidly than it is gained, thus resulting in its load declining over time. In contrast, when plasmid transfer is much higher than plasmid loss, we observed a monotonically increasing relationship opposite to what we initially expected. This represents a scenario where plasmid transfer can compensate for and exceed plasmid loss, thus resulting in its load increasing over time. Finally, when plasmid transfer and loss are balanced, they counteract each other and a unimodal relationship emerges.

Fig. 2:  Effect of the plasmid transfer and loss probabilities on the relationship between the timing of antibiotic administration and  the frequency of transconjugants at the expansion frontier. The relationships were derived from individual-based computational simulations.

So how can we use this information to optimize the administration of antibiotics? Fortunately, the “hit hard and hit quick” paradigm does not generally promote the spread of antibiotic resistance in new genotypes, conditional that plasmid loss is sufficiently low when compared to plasmid transfer. However, for plasmids where the reverse is the case (where plasmid transfer exceeds plasmid loss), then the “hit hard and hit quick” paradigm might indeed have negative consequences that need to be carefully considered. A “hit hard but perhaps wait a while” alternative is similarly harmless in terms antibiotic resistance spread but undesirable if infections can be tackled earlier (hitting fast should always have priority). But again, plasmids that have low loss probabilities will tend to remain longer within the community and be problematic at late antibiotic administration times. Our main point here is that we need a better temporal understanding of the interplay between plasmid properties in complex systems to better understand the problem of antibiotic persistence and spread, and thus develop strategies to tackle the global antibiotic resistance crisis.