Original article: Marta Lukačišinová, Booshini Fernando & Tobias Bollenbach. Highly parallel lab evolution reveals that epistasis can curb the evolution of antibiotic resistance. Nature Communications. Volume 11, Article number: 3105 (2020).
Both infection and cancer are evolutionary diseases where any therapy that falls short of a complete cure exerts evolutionary pressure, driving development of resistance mechanisms. We begin to understand that, like in a game of chess, treating such conditions requires not only a good first move (i.e. administering the right drug), but also anticipating the reaction this and all subsequent interventions might provoke. To facilitate such strategies, the recent study by Lukačišinová and colleagues tracked the evolution of bacteria in response to antibiotics and identified genes that, when lost, diminish the extent of resistance and/or the rate at which it is developed.
The study was made possible by the use of a sophisticated robotic platform in which hundreds of bacterial cultures are monitored and periodically diluted while the antibiotic concentration is individually adjusted – all this in such a way as to keep cells exponentially growing at a rate which is around 50% of what would be the case without the drug. As the evolution progresses, the antibiotic concentration needed to maintain this state – which provides a direct read-out of the IC50 value – rapidly increases, but the trajectories differ between cultures. Still, when starting from the same genetic background, the evolution turns out to be in most cases reproducible both in terms of the observed increase in resistance and mutated genes revealed by sequencing. Bacteria are more predictable than your chess opponent.
In addition to wild-type cells, the authors studied around a hundred different gene deletion strains, some of them selected based on prior knowledge of resistance evolution. When comparing results for different genetic backgrounds, the authors observed an interesting phenomenon that is technically referred to as ‘diminishing-returns epistasis’. Whereas mutants that were already partially resistant underwent a smaller increase in resistance, most cultures that were initially more sensitive could eventually ‘catch up’, reaching a similar upper resistance level.
However, this wasn’t the case for all genetic backgrounds. Most strikingly, the deletion of the gene for TolC, a component of the AcrAB-TolC multi-drug efflux pump, essentially blocked any resistance evolution (for all repeats), and this despite the fact that this strain was initially more sensitive and thus could be expected to show a larger fold change on its way to the upper resistance limit. One could imagine that this would imply that resistance evolution in other strains necessarily progresses through overexpression or gain-of-function mutations in this particular gene. However, this is not the case. The evolution of resistance might involve mutations in other genes (typically mutations in three to four genes were fixed in the course of the experiment), but the beneficial effects of these mutations do depend, at least initially, on TolC, preventing the changes from getting fixed in the tolC-deficient strain. In technical terms, these resistance mutations show ‘epistatic interactions’ with tolC, even if they do not directly affect tolC. An interesting result was observed for another component of the efflux pump complex, ArcA, where four out of five repeats showed no resistance, while one repeat escaped the evolutionary dead end by overexpressing AcrEF-TolC, a rarely used alternative to AcrAB-TolC. (I guess one could compare this result to a particularly difficult chess puzzle, which one in five players manages, nonetheless, to solve by making this one correct move). These results generally show that efflux pumps, which have long been implicated in antibiotic resistance (by literally pumping drugs out of the cell), are also important for the evolvability of that resistance – and thus make a particularly good target for adjuvants that could aid antimicrobial therapy.
Finally, it is worth mentioning that the results for mutations in chaperones such as DnaK (the bacterial homologue of Hsp70), which showed decreased rate of resistance development, bring to mind the model put forward by the late Susan Lindquist and her colleagues of chaperones as 'buffers' or ‘capacitors of evolution’. According to this model, which was based on insights from evolutionary biology but also rapid evolution of cancer cells, chaperones (i.e. proteins that help other proteins fold into a proper shape and prevent them from aggregating) can facilitate evolution because mutations that arise in the course of evolutionary transitions can destabilise protein folds – but this destabilisation can in turn be counteracted by chaperones. I guess this does not have to be limited to point mutations; also mutations that result in overproduction (which generally puts more pressure on the folding machinery) or loss of a protein (which can lead to destabilization of its interaction partners, especially if they form a tight complex) could in theory be more easily tolerated in the presence of chaperones. The results by Lukačišinová and colleagues lend support to such models. I know of experimental therapies that aim to target chaperones in cancer, but it might also be an interesting idea for antimicrobial interventions.
The study discussed here was performed in a model organism (E. coli) by using a limited set of genetic backgrounds, and the approach was by necessity reductive. For example, cultures were grown in isolation from the host and other bacteria, thus preventing horizontal transfer – a phenomenon that is known to play a major role in antibiotic resistance. Nonetheless, thanks to an ingenious experimental design and sophisticated technology the study comes close to a real-life scenario of a ‘game of chess’ between bacteria and their human hosts, providing both specific and general (conceptual) insights into how we might go about winning this game.
Link to the cover photo: https://www.flickr.com/photos/nemodus/4980060229/